Hip Fracture Patients at Risk of Second Hip Fracture: A Nationwide Population-Based Cohort Study of 169,145 Cases During 1977–2001

Authors


  • Dr. Ryg has received speaker's fee or paid traveling expenses from Eli Lilly, MSD, Nycomed, Roche, and Servier. Dr. Overgaard has received consultancy fees from Eli Lilly and MSD. Dr. Brixen has received consultancy fees from Servier, Novartis, Eli Lilly, Nycomed, and Osteologix, as well as speaker's fees from Eli Lilly, Novartis, and research grants from MSD. All other authors state that they have no conflicts of interest

  • Published online on February 16, 2009

Abstract

In patients with prior hip fracture (HFx), little is known about time frame and risk factors of second HFx, as well as the ensuing mortality. The aim of the study was to elucidate the incidence of second HFx and subsequent mortality. All 169,145 patients with a first HFx in Denmark during 1977–2001 were followed for up to 25 yr and compared with the background population. Data on fractures, vital status, comorbidity, redeemed prescriptions, and socio-demographic variables were retrieved from national registers. Median follow-up was 3.8 yr, corresponding to 1,041,177 patient-years. A total of 27,834 patients had a second HFx. The cumulative incidence was 9% after 1 yr and 20% after 5 yr, being significantly higher than expected (2% and 12%, respectively; p < 0.05). The RR of second HFx was 2.2 (95% CI: 2.0–2.5) at 1 yr and did not normalize until 15 yr (RR = 1.01, 95% CI: 1.0–1.02). Risk factors for a second HFx were female sex (HR = 1.36, 95% CI: 1.32–1.40), age (HR = 1.68, 95% CI: 1.60–1.76 in patients >85 yr), alcoholism (HR = 1.61, 95% CI: 1.51–1.72), any prior fracture (HR = 1.08, 95%CI :1.04–1.11), and living alone (HR = 1.06, 95% CI: 1.04–1.09). Both sexes had higher mortality at 1 and 5 yr after a second HFx compared with the background population (men—1 yr: 27% versus 9%, p < 0.05; 5 yr: 64% versus 40%, p < 0.05; women—1 yr: 21% versus 10%, p < 0.05; 5 yr: 58% versus 41%, p < 0.05). Patients with HFx are at 2-fold risk of further HFx and the subsequent mortality is highly increased. We propose that programs for secondary prevention should be developed and tested.

INTRODUCTION

Lifetime risk of hip fracture (HFx) at the age of 50 yr is ∼11–23%(1–4) and 3–11%(2–4) in women and men, respectively, with great variations worldwide.(5) Although, some studies have found the age- and sex-specific incidence of HFx to increase,(6–8) recent studies have reported the increase to level off(9) or even decrease.(10,11) Nevertheless, the prevalence of HFxs is expected to keep rising because of the increasing number of elderly.(12,13) Of all fracture types occurring in the elderly, HFx is considered to have the most severe impact leading to a marked decrease in quality of life,(14) dependency of walking aids or help from others,(15,16) and excess mortality.(17–19) Also, the burden on the society is considerable because of the direct costs of hospital care, rehabilitation, and nursing home facilities. HFxs are estimated to account for 72% of the total costs of fractures.(20) Most HFxs occur late in life because of a combination of falls(21) and low BMD.(22–25)

Despite this, only few HFx patients are systematically offered secondary fracture prevention.(26–30) This may partly be because of the perception that intervention is meaningless because of lack of evidence based therapy, high age, mortality or comorbidity, or low incidence of further HFxs.

Several randomized, placebo-controlled trials on primary fracture prevention using calcium plus vitamin D supplementation(31–33) or hormone therapy(34,35) have shown significant reduction of HFx incidence. Similarly, alendronate,(36) risedronate,(37) and zoledronic acid(38) have been shown to reduce the incidence of HFx in patients with osteoporosis. Also, a population-based, retrospective study recently reported a significantly reduced rate of recurrent HFx in patients, redeeming antiresorptive agents (adjusted hazard ratio [HR], 0.74; 95% CI: 0.64–0.86).(39) Although the incidence of HFx increases with age(2,4,29) and the mean age of HFx patients in Denmark is 77.0 ± 13.0 yr,(40) the mean survival time of the Danish background population at that age is still 11.1 yr and that of HFx patients is 6.2 yr.(40) These data are corroborated by an Australian study comparing life expectancy after HFx with the background population.(17) At the age of 75–79 yr, the life expectancy was 9.4 and 12.0 yr in men and women, respectively, whereas that of HFx patients was 4.8 and 10.8 yr. The mortality after the first HFx is indeed increased. A recent Danish case-control study(40) reported an excess mortality of 19% during the first year after HFx, but the excess mortality stabilized and was only 2% per year thereafter. A number of case-control studies have reported that patients with HFx are more often affected by comorbidity than the background population.(18,40) Also, a recent cohort study found a significant increase in hospital mortality with increasing comorbidity,(41) but long-term mortality seems unrelated to the prefracture comorbidity.(40) However, the impact of a weighted index of comorbidity(42) on HFx recurrences and subsequent mortality is unknown. Even though a number of reports have shown that any prior fracture increases the risk of first HFx,(43–47) fewer studies have investigated the risk of a second HFx. These studies have relied on self-reported data on fractures,(48,49) were limited to women,(49,50) had a short (≤4 yr) follow-up time,(51,52) comprised only a few (<300) events,(48,50–57) or included no data on comorbidity.(49,51,54,55,58) Moreover, only a single study with a small number of events (n = 71) explored the mortality after second HFx.(53)

We studied the incidence of second HFx, the ensuing mortality, and the possible impact of comorbidity in a nationwide, population-based cohort of 169,145 male and female HFx patients during 1977–2001.

MATERIALS AND METHODS

Study design

The study was designed as a nationwide, population-based, historical cohort comprising all patients with HFx defined as femoral neck and trochanteric fractures. Primary outcome was second HFx. Secondary outcome was mortality after the second HFx.

Sources of data

Population-based registers in Denmark offer good possibilities for studies on the occurrence of fractures.(59) The National Hospital Discharge Register covers all in-patient records in Danish hospitals (excluding departments of psychiatry).(60) The register was founded in 1977 and since 1995 also comprises out-patient records. Records are available for any given International Classification of Diseases (ICD) code or surgical procedure. The register has a high validity(61) regarding the diagnosis of fractures with a previously reported precision of 97%.(62) The Psychiatric Central Register covers in- and out-patient records from all Danish mental hospitals since 1968.(63) This register has a high validity as well.(64) The National Prescription Database is run by the Danish Medicines Agency (www.dkma.dk). Since January 1, 1996, all redeemed prescriptions in Denmark have been registered in the database. Data include Anatomical Therapeutic Chemical Classification System (ATC) codes, dose of the drug, and package size; however, no data are available on the actually prescribed dosage. The National Bureau of Statistics provides detailed statistical information on the Danish Society including socio-demographical variables (www.dst.dk/HomeUK.aspx).

Linking of data

Since 1968, all Danish citizens have been assigned a unique personal 10-digit Civil Registration number. The Ministry of Interior Affairs and Health's Department are in charge of the Danish Civil Registration System administrating basic personal information such as vital status and residence (www.cpr.dk). The civil registration number provides means for valid linkage of information at the individual level between all the population-based registries.

Identification of HFx cohort

All patients with HFx (International Classification of Diseases, Eighth Revision [ICD-8] codes 820.00 and 820.01 and ICD-10 codes S72.0 and S72.1) in Denmark in the 25-yr period between January 1, 1977 and December 31, 2001 were retrieved from the National Hospital Discharge Register and comprised our cohort (index HFx). ICD-9 codes were not used because these were never implemented in Denmark. The fracture cases were further validated using the surgical procedure codes of primary hip arthroplasty and osteosynthesis of hip fractures according to the classification of Surgical Procedures from the National Board of Health (75636, 75637, 75638, 7003x [x = 0–9], 7043x [x = 0–9], 7053x [x = 0–9], 7063x [x = 0–9], NFBxy [x = 0–9, y = 0, 1, or 2], NFJxy [x = 0–9, y = 0, 1, or 2]).

Identification of HFxs during follow-up

Similarly, we retrieved the occurrence of second HFx from the National Hospital Register appearing in Denmark in the period between January 1, 1977 and December 31, 2001. The fractures where validated by matching fracture date with date of surgical procedure codes, allowing us to exclude records with wrong procedure codes like reoperation caused by fracture complication. Furthermore, we excluded patients referred from out-patient clinics because these most likely were hospitalized because of fracture complications rather than a new fracture. Finally, we excluded records entered while the patient still was in the hospital after the index HFx.

Pharmaceutical treatment of osteoporosis

Data on anti-osteoporotic medication (i.e., hormone therapy [HT], fluoride, etidronate, alendronate, risedronate, raloxifene, and calcitonin) were retrieved from the National Prescription Database. We have no data on calcium or vitamin D supplementation because they are not reimbursed.

Comorbidity and socio-demographic variables

The Charlson Index, which is a validated index of 19 medical conditions, was used to express a weighted index of comorbidity(42) on the basis of data retrieved from the National Hospital Discharge Register. Alcoholism was defined as either a previous prescription of disulfiram in the National Prescription Database or a prior diagnosis of alcoholism in the National Hospital Discharge Register or the Psychiatric Central Register. Socio-demographical variables on income and civil status were retrieved from the National Bureau of Statistics for the year of fracture.

Vital status

Information on possible migrations or date of death was retrieved from the Central Person Register. Detailed information on the same cohort on mortality after first HFx has previously been reported.(40)

Background population

We compared the fracture incidences in the HFx cohort with that of the entire Danish population stratified by age and sex using data from the National Hospital Discharge Register on all fractures in Denmark between 1977 and 1999 and census data from the National Bureau of Statistics. The expected incidence of HFx was given as the incidence of first HFx in the background population. The Danish population comprises ∼5.5 million subjects.

Ethics

This study was subject to control by the National Board of Health and the Danish Data Protection Agency. It was reported according to international guidelines for reporting observational studies (STROBE).(65)

Statistics

We included data from the establishing of The National Hospital Discharge Register and 25 yr ahead to have valid data from the whole time period. Because data are available from the registers on addresses of all individuals, we were able to account for any migration. Data are shown as mean ± SD, median [range], or HR (95% CI). Comparison between groups was performed using χ2, Student's t-test, or Wilcoxon's test, as appropriate. The cumulative incidence of second HFx was calculated using Kaplan-Meier survival analysis. Cox' proportional hazard model was applied to assess the relationship between covariates and risk of second HFx. Analyses were performed using STATA 8.1 (STATA, College Station, TX, USA) and SPSS 14.0 (SPSS, Chicago, IL, USA)—both in the Unix version. p values <0.05 were considered statistical significant.

RESULTS

Demographics

In the study period, the Danish population totaled 5,080,000 in 1977 and 5,368,000 in 2001. The mean percentage of women in the population during the total period was 50.66 ± 0.07%. The percentage of individuals 0–49, 50–64, and 65+ yr of age was 67.7, 16.8, and 15.5 versus 64.1, 18.9, and 17.0 in women and 71.9, 16.2, and 11.9 versus 68.0, 19.4, and 12.6 in men in 1977 and 2001, respectively.

Baseline data

Baseline data on the 169,145 HFx patients are shown in Table 1. None of these were lost to follow-up because of migration. The majority of patients were women (72.4%, p < 0.01), being older than men (78.6 ± 11.1 versus 72.8 ± 16.3 yr, p < 0.01). Women had the highest prevalence of any fracture before the index HFx (21.7% versus 17.3%, p < 0.01). Comorbidity as defined by a Charlson index >0 was most frequent in men compared with women (50.4% and 43.3%, p < 0.01). Only 0.5% and 3.5% (p < 0.01) of the male and female patients, respectively, had redeemed any prescriptions of osteoporosis therapy in the time period with available data from The National Prescription Database (1996–2001). In our study, a total of 6833 (4.0%) patients were <50 yr of age. Excluding patients <50 yr of age did not alter the results.

Table Table 1.. Baseline Characteristics of the Danish Hip Fracture Cohort
original image

HFxs during follow-up

Median follow-up in the total cohort was 3.8 yr [0–25 yr], corresponding to 1,041,177 patient-years (py). A total of 27,834 patients (5,369 men, 22,465 women) had a second HFx during the study period. The overall incidence of second HFx was 39 per 1000 py, being highest in women (42 per 1000 py in women versus 31 per 1000 py in men, p < 0.05). The age- and sex-specific incidences per 1000 py in age groups <50, 50–59, 60–69, 70–79, 80–89, and ≥90 yr were 24, 34, 40, 51, 62, and 73 in women and 14, 26, 30, 43, 58, and 68 in men. The increasing incidences seen with age were significant in both sexes.

The incidence of second HFx was significantly increased in patients with HFx compared with the age- and sex-matched background population (Fig. 1). After 1 yr, the cumulative incidence of a second HFx was 9% and 2%, in the HFx cohort and controls, respectively (p < 0.05). After 5 yr, patients with prior HFx still had significantly increased risk of new HFx compared with controls (20% versus 12%, p < 0.05), whereas the increased risk leveled of over time with no significant difference seen after 15 yr. Stratifying data according to sex did not change the results. The relative risk (RR) of a second HFx was increased to 11.8 (95% CI: 11.2–12.5) 1 mo after the index fracture and 2.2 (95% CI: 2.0–2.5) at 1 yr, but was normalized to 1.01 (95% CI: 1.0–1.02) after 15 yr (Fig. 2). When stratifying for age or comorbidity, RR did not change significantly (data not shown).

Figure Figure 1.

Risk of second HFx compared with the risk of HFx in an age- and sex-matched normal population. Calculated as actuarial survival.(40) Because of the large cohorts, CIs were too small to be presented on the curves.

Figure Figure 2.

RR of second HFx compared with the expected risk in an age- and sex-matched normal population in selected time intervals after the first HFx.

Adjusting for mortality showed a significantly higher 5-yr HFx-free period in the younger patients in both sexes and highest in men (Table 2). The risk of second HFx increased with increasing age. The percentages of patients at 50–59 yr of age with a second HFx after 5 yr were 14.3 ± 0.7% and 19.4 ± 0.6% in men and women, respectively. The corresponding percentages in patients 80–89 yr of age were 19.1 ± 0.6% and 23.3 ± 0.3% (Table 2).

Table Table 2.. Hip Fracture Free 5-yr Survival After the First Hip Fracture in Men and Women Stratified According to Age
original image

Using a Cox' proportional hazard model, female sex (HR = 1.36, 95% CI: 1.32–1.40), age (HR = 1.68, 95% CI: 1.60–1.76 in patients >85 yr), alcoholism (HR = 1.61, 95% CI: 1.51–1.72), any prior fracture (HR = 1.08, 95% CI: 1.04–1.11), living alone (HR = 1.06, 95% CI: 1.04–1.09), and having a higher income (HR = 1.03, 95% CI: 1.00–1.05) were all identified as significant risk factors for second HFx (Table 3). Charlson index was not a statistically significant predictor of the risk of second HFx.

Table Table 3.. Risk Factors for a Second Hip Fracture Identified Using Cox' Proportional Hazard Model
original image

Mortality

A total of 121,953 (72.1%) HFx patients died during follow-up. Both 1- and 5-yr mortality after second HFx was significantly higher in men compared with women (27% versus 21%, p < 0.05, and 64% versus 58%, p < 0.05, respectively). Furthermore, both sexes showed a significantly higher mortality at 1 and 5 yr after second HFx compared with the mortality of the background population (men—1 yr: 27% versus 9%, p < 0.05; 5 yr: 64% versus 40%, p < 0.05; women—1 yr: 21% versus 10%, p < 0.05; 5 yr: 58% versus 41%, p < 0.05; Fig. 3). A significantly increased mortality after the second HFx was seen with age (data not shown). Finally, no significant increased mortality after second HFx was seen with increasing Charlson index.

Figure Figure 3.

Mortality after second HFx stratified according to sex compared with the mortality of the background population. Because of the large cohorts, CIs were too small to be presented on the curves.

DISCUSSION

Our study showed a high incidence of second HFx in both men and women, with a 5-yr cumulative risk ranging from 9.5% to 20.4% in men and 15.9% to 23.3% in women depending on age. This was approximately twice the incidence observed in the background population. We also observed a significantly increased mortality after second HFx compared with the background population. Thus, 1-yr mortality was increased 2-fold in women and 3-fold in men.

The high incidence rate of second HFx reported in our study corresponded well with previously reported data.(50,52,53,55–57) In our study, the risk of second HFx was increased almost 12 times at 1 mo and more than doubled at 1 yr; however, the majority of second fractures occurred at a later time, and the risk remained significantly elevated until 15 yr of follow-up. In the study by Melton et al.,(54) the overall risk was also increased compared with controls. The risk was somewhat <2-fold but it did not level off over time. This might in part be because of limited power, because the study only reported few second HFx occurring in the first 18 mo (n = 25).

In the Cox' proportional hazard model, HFx recurrence was predicted by age, prior fracture, female sex, alcoholism, and living alone. The 5-yr HFx risk was significantly increased with age in both sexes. This is in accordance with some previous studies,(53,58) whereas other studies could not show such interaction, probably because of a low number of events (n = 34–53).(50,52,57) One study reported a decreased risk of second HFx with increasing age.(54) The reason for this is obscure. We found a high prevalence of prior fractures in both women (21.7%) and men (17.3%), increasing the risk of second HFx significantly by a factor 1.08. In the study by Chapurlat et al.,(50) they did not find an increased risk. This may be attributable to limited power or to the examination of a healthier cohort because they only included community-dwelling subjects. Despite this, our data indicate a possible underestimation of the impact of prior fracture because the Danish National Hospital Discharge Register did not include out-patient visits until 1995. In our study, female sex was a strong predictor for second HFx. Melton et al.(54) reported similar data; however, they also found that the effect of sex diminished when controlling for trauma severity. We have no data on the trauma mechanism and thus cannot elucidate this further. Other studies reporting on sex differences did not find an increased risk in women. This may partly be explained by the few (n = 6–25) male second HFxs reported in those studies.(51–53,56,57) High intake of alcohol increases the risk of HFx.(66,67) In our study, alcoholism was defined as a prescription of disulfiram or prior hospital diagnosis of alcoholism. This approach may identify patients with the highest alcohol intake only, thereby underestimating the true incidence of alcoholism. However, alcoholism as defined in our study, was associated with a significantly increased risk of recurring HFx. Because of the register based nature of our study, we cannot analyze any dose-response relationship or threshold effects. Only a single study has addressed this issue previously. Chapurlat et al.(50) found no association between self-reported alcohol intake and risk of second HFx in their study on women. The event of second HFx in this study, however, was low (n = 53). Finally, we found that patients living alone had a higher risk of second HFx. In the study by Wolinsky and Fitzgerald,(56) they did not find association but had very few cases (n = 27).

As indicated, most of the other studies investigating the risk of second HFx only had small number of events, which may have had an impact on their results because of limited power.(50–54,56,57) Because of the large size of this study, our estimate of the RR associated with the above mentioned factors, however, seems solid, with narrow confidence limits.

It has been well documented that HFxs are mostly seen in the elderly and that incidence increases with age.(2,4,29) Less is known about younger patients with HFx. Previously, a Norwegian(68) and Danish(69) study reported a percentage of HFx patients 20–50 yr of age, in the range of 2.4–2.6, with an incidence of 12–13 per 100.000 py. The Norwegian group has recently reported that the majority of HFxs in young patients (i.e., 20–50 yr of age) are caused by low-energy trauma. Furthermore, they found lower BMD values regardless of trauma mechanism in these patients compared with a reference population.(70) Our study was nationwide, covering all HFxs at any age. Even though only a few patients were <50 yr of age in our study and despite that this age group had the lowest 5-yr cumulative risk of second HFx, the risk was still increased compared with the expected risk. Because of our study design, we have no BMD data to reproduce the Norwegian findings, but the results indicate that future observational studies or studies on fracture prevention should consider including this age group as well.

The time interval from first to second HFx is interesting from a clinical point of view; however, this parameter can only be defined directly in a cohort of patients followed lifelong. Another way of addressing this issue, however, is to report the cumulative incidence at discrete points in time. In our study, the cumulative incidence of second HFx was 9% after 1 yr and 20% after 5 yr. Two population-based American studies showed almost identical data with a cumulative incidence of second HFx of 1–2.5% after 1 yr and 8–8.2% after 5 yr,(53,54) whereas a Finish study(52) reported higher incidences (5.1% at 1 yr and 8.1% at 2 yr). These differences may mirror the large variation in risk of HFx and life expectancy in different populations,(5) but data from all four studies suggest that there is ample time to initiate preventive measures in the majority of patients.

The majority of our patients had concomitant diseases as indicated by a Charlson index >0. Similar findings were reported by Leibson et al.,(18) who found that 45% of the HFx patients had a Charlson index ≥1. In both studies, comorbidity was most frequent in men. The comorbidity index did not predict the risk of new HFx in our series. However, the Charlson index was originally designed to predict mortality. Moreover, some important HFx risk factors (e.g., dementia) are mostly diagnosed in primary care or out-patient clinics and thus not fully incorporated in our data. Indeed, a Japanese study(57) reported significantly increased risk of second HFx in patients with dementia (OR, 3.07; 95% CI, 1.58–5.96). Other studies have used different indices of comorbidity, with divergent results. Lönnruss et al.(52) found no significant risk of second HFx with any of a number of different comorbidities including dementia, whereas Berry et al.(53) reported a significantly increased risk in patients with a high level of functioning compared with those with moderate functioning (HR, 2.7; 95% CI, 1.1–6.9). Finally, Wolinsky and Fitzgerald(56) reported a significantly increased risk in patients with poor perceived health status (RR, 2.21; 95% CI, 1.03–4.72). Similarly, Chapurlat et al.(50) found that walking for exercise, which is an indirect marker for functional status, was a protective predictor (RR, 0.5; 95% CI, 0.3–0.9). The explanation for these discrepancies might be caused by differences in defining comorbidity used in the varying studies.

We observed that patients with a second HFx had a substantial mortality of 27% and 64% in men and 21% and 58% in women after 1 and 5 yr, respectively. This was virtually similar to the previously published mortality observed in the cohort after first HFx (32.9% versus 63.4% in men and 23.9% versus 56.2% in women).(40,71) In both instances, 1- and 5-yr mortality was significantly highest in men. Only the study by Berry et al.(53) has previously explored the mortality after second HFx, and despite an impressive follow-up time (51 yr), the actual number of events in that study was small (n = 71). They reported an overall 1- and 5-yr mortality of 15.9% versus 45.4% and 24.1% versus 66.5% after first and second HFx, respectively.(53) No data were available on possible differences seen between sexes after second HFx. The discrepancy between studies, whether mortality was increased after second compared with first HFx, is caused by higher mortality after first HFx in our study, because both studies reported similar high mortality rates after second HFx. A recent study found no association between the preexisting comorbidity as expressed by the Charlson index and the increased mortality seen after first hip fracture.(40) Interestingly, this finding was reproduced in this study, because no increased mortality after second HFx was seen with increasing index. Our findings indicate that the increased mortality seen after second hip fracture may be caused by postfracture conditions or frailty factors not included in the index. However, course of death after second HFx is beyond the scope of this paper and we have no data to explore this further.

Our study had some limitations. First, we estimated the risk of a second HFx using survival analysis. Because mortality is increased after HFx, this approach underestimates the true incidence. Second, our data did not allow us to account for patients who had prosthetic surgery in the hip (e.g., caused by osteoarthritis and thus only had one hip at risk). Nor could we account for patients who sustained a HFx before 1977 and thus wrongly had their second HFx classified as an index HFx while having no hips at risk. These biases, however, would lead to an underestimation of the true incidence of second HFx. Third, we did not stratify our data according to trauma mechanism. A recent study(72) using data from the prospective cohorts of the Study of Osteoporotic Fractures (SOF) and the Osteoporotic Fractures in Men Study (MrOS) reported equally increased risk of future fractures in patients with high- or low-trauma nonspine fractures. However, the study did not report specifically on second HFx. Indeed, an early, population-based study from Melton et al.(54) found that patients with HFx caused by high-energy traumas did not have an increased risk of second HFx. The possible bias introduced by this lack of stratification in our study would also seem to underestimate the incidence of second HFx. Also, we did not present data on fracture type (trochanteric or neck) because a previous study showed that classification of HFx subtypes in the Danish National Hospital Register is not completely reliable.(55) Fourth, instead of being registered as post-HFx complications, some of these incidents (e.g., infections, periprostetic fracture, alloplastic surgery, or removal of osteosynthesis material) may have been misclassified as new HFxs in the National Hospital Discharge Register. This bias would overestimate the true incidence of second HFx and is crucial for our data. To omit registrations not caused by new HFxs, we only included patients with matching fracture and surgical codes. Furthermore, patients referred to the hospital for fracture revision from an out-patient clinic and events registered during the patients in-hospital stay after an index HFx were not included because these records most likely are caused by fracture complications. However, this also excludes possible fractures sustained when patients fell out of bed during hospitalization. The possible bias introduced by this would underestimate the fracture incidence, but numbers are small and we have no means to explore this further. Validation is a potential weakness in all register-based studies. However, a smaller-scale study using data from the Danish National Hospital Discharge Register showed a high validity regarding fractures in general.(62) By making several qualitative approaches, our efforts have been to use as valid data as possible. Unfortunately, we had no possibility to validate our data further. Finally, because our study was register based, we cannot address the importance of BMD, hip structure, anthropometrics, or lifestyle.

Our study, however, also had important strengths. First, it was population based and nationwide, comprising all HFxs for a period of 25 yr. The number index cases (169,145), second HFxs (27,834), and follow-up time (25 yr corresponding to >1 million py) were substantial, and our study was larger than all previous studies combined. Second, the existence of the extensive Danish registers allowed us to collect valid data on deaths, redemption of prescriptions, comorbidity, and socio-demographical variables. Furthermore, we were able to compare most of the findings with age- and sex-matched data from the general population using the same registers. More than 70% of the patients were followed until death during this long follow-up, and no patients were lost to follow-up because of immigration.

We conclude that HFx patients have a 2-fold risk of a second HFx. Although many second HFxs occurred shortly after the first, a large proportion happened >1 yr after the first HFx. Alcoholism, living alone, female sex, age, and any prior fracture are significant risk factors for a second HFx. Mortality after second HFx is increased to the same magnitude as after first HFx but seems unrelated to preexisting comorbidity. In light of the worldwide rise in the proportion of elderly at risk, our data suggest that programs of secondary prevention of HFxs should be developed and tested.

Acknowledgements

The authors thank University of Southern Denmark, Danish Centre for Evaluation and Health Technology Assessment, the Christenson-Ceson's Family Foundation, and Helga and Peter Korning's Foundation for financial support. Statistics Denmark (Danmarks Statistik) is acknowledged for invaluable help with the data.

Ancillary