The authors state that they have no conflicts of interest.
Long-Term Survival and Fracture Risk After Hip Fracture: A 22-Year Follow-Up in Women†
Article first published online: 30 JUN 2008
Copyright © 2008 ASBMR
Journal of Bone and Mineral Research
Volume 23, Issue 11, pages 1832–1841, November 2008
How to Cite
von Friesendorff, M., Besjakov, J. and Åkesson, K. (2008), Long-Term Survival and Fracture Risk After Hip Fracture: A 22-Year Follow-Up in Women. J Bone Miner Res, 23: 1832–1841. doi: 10.1359/jbmr.080606
- Issue published online: 4 DEC 2009
- Article first published online: 30 JUN 2008
- Manuscript Accepted: 27 JUN 2008
- Manuscript Revised: 18 JUN 2008
- Manuscript Received: 6 MAR 2008
- fracture risk;
- hip fracture;
- residual lifetime risk;
Hip fracture is associated with high early mortality. Little is known about long-term survival and subsequent fracture risk. The aim of this study was to evaluate survival and fracture risk after hip fracture in women at different ages. All women suffering a hip fracture during 1984–1985 in Malmö, Sweden, were identified (n = 766) and followed up to 22 yr or death. All new radiographic examinations related to musculoskeletal trauma with or without fracture were registered. Survival (mortality) and fracture was evaluated in 5-yr age bands and in age groups (<75, 75–84, and ≥85 yr). Mean age was 79.6 ± 9.9 yr (range, 31.6–99.4 yr), with 42% between 75 and 85 yr of age. Overall 22-yr survival was 6%: 79% at 1 yr, 48% at 5 yr, and 33% at 10 yr (i.e., population at risk). One-year mortality was 7%, 21%, and 33% for <75, 75–84, and ≥85 yr of age, respectively, and 95% of those ≥85 yr were dead at 10 yr. Prior hip fracture did not affect age-adjusted mortality (OR, 1.05; 95% CI, 0.756–1.20; p = 0.15). A total of 768 fractures were registered at 715 occasions in 342 women (45%; mean, 2.3 fractures/woman; range, 1–11 fractures/woman). Of the fracture occasions, 15% occurred within the first year, 27% within 2 yr, and 73% within 5 yr. The residual lifetime fracture risk was 45%, with a mortality-adjusted increase to 86%. The 10-yr fracture risk was 40%; with a mortality-adjusted increased to 65%. In conclusion, almost one half of all women with a hip fracture suffer a new fracture during their remaining lifetime. Fracture risk is highly dependent on age and survival, emphasizing that preventive strategies need to be tailored to each age group specifically.
Hip fracture is the most devastating outcome of osteoporosis and is estimated to increase globally within the first one half of this century to reach >6 million by 2050.[1-3]
Hip fracture commonly affects an already frail population, with the hip fracture either as the final fracture event in a person's life or as one in a series of fragility fractures over a lifetime. Subsequently, cost-effective programs to prevent future fractures in hip fracture patients need consider, not only the overall risk increase related to a prior fracture, but also estimates of short- and long-term survival.
Hip fracture is associated with increased short-term mortality, particularly within the first 3–6 mo,[4-6] whereafter it seems to plateau. One-year mortality varies between 11% and 34% according to previous reports.[7-11] The short-term mortality may be directly attributable to the fracture, to the advanced age of the hip fracture patient, and to co-existing morbidity from other organ systems.[11-16]
The risk of subsequent fracture is influenced by survival, which is statistically dependent on age. In a study using a 5-yr observation period, the estimated fracture risk after an osteoporotic fracture seemed to be greater during the first year, but also in those at age 60 compared with those age 80 over the entire observation period, taking mortality into account. Furthermore, a low-energy fracture sustained in middle-aged persons predisposes for a future fracture,[23-25] and the subsequent risk also includes a second hip fracture. However, estimates of subsequent fracture risk in hip fracture populations are mainly based on statistical modeling from short-term data,[27-30] with only one study relying on long-term observations. This means that most studies are neither reporting all fractures nor able to use extended exposure time based on residual lifetime.
An additional factor influencing fracture risk is the propensity to fall. With the exception of vertebral fractures, the majority of osteoporotic fractures are fall related. Falls are reported annually in 30% of those 65 yr of age and in 50% of those ≥80 yr of age.[32, 33] However, it is estimated that only 1–14% of all falls lead to a fracture, with a higher proportion among the oldest.[34-36] Virtually all hip fractures are related to a fall; hence, hip fracture populations are likely to be at even higher risk of falling and subsequent fractures. The documentation of falls is, however, complex, and for this reason, it is difficult to evaluate the risk exposure; nevertheless, it is important in terms of prevention.
A better understanding of the long-term relationship between hip fracture, survival after hip fracture, and the risk of subsequent fractures in women is needed to ensure effective targeting of fracture intervention programs. This study, using a population-based cohort of women with hip fracture, is designed to address these questions: (1) what is the influence of age on long-term survival, (2) how many subsequent fractures occur and in what proportion of the patients; (3) what is the residual lifetime and the 10-yr fracture risk; and (4) how does trauma, requiring hospital attention, correlate with fracture risk. To this end, we identified all female hip fracture patients during 1984–1985 and followed the cohort up to 22 yr or until death, using radiographic examinations related to musculoskeletal fracture and trauma as measures of outcome.
MATERIALS AND METHODS
All women suffering a hip fracture (=index fracture) during 2 consecutive yr, 1984 and 1985, were identified through the database of the Department of Radiology at Malmö University Hospital, Malmö, Sweden, a department that has saved the radiographic records and X-rays since the beginning of last century. The Department of Orthopedics at Malmö University Hospital is the only unit treating adult fractures in a well-defined catchment area (260,000 inhabitants).
In this study, we chose to include all adult women with hip fracture (>20 yr), because we wished to capture the entire fracture group and stratify for age. In all, 778 female patients with a hip fracture were identified through the database; two patients had emigrated at unknown dates during follow-up and were subsequently excluded. After excluding fractures caused by high-energy trauma (n = 6) and pathological fractures (n = 4), 766 remained in the study. Of these, one patient emigrated and was followed until date of emigration (t = 12.5 yr). The patients were followed up to 22 yr (i.e., until death or April 2006).
All musculoskeletal radiographic examinations occurring from the entry date were individually assessed for each woman, and data were exclusively collected by one person (MF) (Fig. 1).
The date of death was identified through the Epidemiological Center, National Board for Health and Welfare, Sweden. The study received ethical approval from the Institutional Review Board, Lund University, Lund, Sweden.
Hip fracture was defined as a fracture of the proximal femur ranging from cervical neck (intracapsular) to the subtrochanteric region. The date of the index fracture was registered, as was side (left/right), type of fracture, cause of trauma, type of treatment, and presence of co-existing fracture. Information on the index fracture was cross-validated with the surgical register and the medical charts. Co-existing fracture means a fracture at another site, caused by the same trauma and occurring simultaneously with the index hip fracture. In addition, hip fractures having occurred before the index hip fracture were recorded.
Musculoskeletal radiographic examinations
After the index fracture, every musculoskeletal trauma leading to X-ray of a skeletal structure or skeletal region was recorded. The following information was documented at each occasion: date, type of trauma, number of structures examined, and if fractured, type of fractures. When subsequent X-rays, within days or a few weeks, of an original and previously registered trauma were performed, the newly examined structure was added to this original trauma occasion. Discovery of a new fracture was similarly added to the original trauma occasion.
An X-ray occasion was defined as the day of X-ray caused by a specific trauma occasion, whereas X-ray examination refers to the specific structure or region examined, for instance, forearm, spine, or ankle. Structures close to each other, for example, hip-pelvis, forearm-wrist, and shoulder-humerus, were considered one X-ray examination. Repeat X-ray examinations related to fracture follow-up were excluded to avoid overestimation of trauma.
The description of trauma is based on the information from the attending doctor on the X-ray referral sheet on the referral for X-ray. Spinal radiograms performed because of suspected vertebral fractures were also included. Musculoskeletal trauma was coded as follows: low-energy trauma (including microtrauma, fall from a standing height, a chair, or a bed); high-energy trauma (traffic accident or fall from a building); and tumors or similar (pathological fracture). This classification was also used when evaluating the cause of X-ray examination and fracture at follow-up.
Baseline descriptive data are presented as mean, SD, and range. Baseline data are reported in 5-yr age bands; those <50 yr of age (n = 9) are reported as one age band. When comparing groups of patients, three age groups were defined using the following stratification: <75, 75–84 yr, and ≥85 yr at index hip fracture. Age was stratified on the basis of mean age at the time of the index hip fracture ±5 yr. Student's independent t-test was used to identify difference in mean age between groups. Cox proportional hazard regression was used to adjust for age difference.
Survival of women after the index hip fracture was quantified using survival function estimates by the Kaplan-Meier method. Differences in survival, based on Kaplan-Maier survival curves, were tested using the log-rank test (or Gerhan-Wilcoxon test). The censoring event was to be alive at the end of follow-up, and a complete event was mortality date. Fracture risk was quantified using the same method; being fracture free at the end of follow-up was the censoring event, and the date of a first new fracture was a complete event. Adjustment for mortality was made by also using date of death as censor.
Fracture incidence after the index hip fracture is reported in percentage by age group. To clarify comparison with other studies with varied follow-up, fracture incidence is also overall reported per 1000 person-years and calculated by dividing the number of women with fracture with the sum of years for all women from index hip fracture to first fracture, death, or end of follow-up. The level of significance was set at p < 0.05. The data were analyzed using STATISTICA software, release 7.1.
The mean age at the time of the index hip fracture was 79.6 ± 9.9 yr (range, 31.6–99.4 yr). Excluding women <50 yr of age (n = 9, two <40 yr of age) only slightly altered the mean age (80.0 ± 9.1 yr; range, 52.9–99.4 yr). The majority of the women were ≥75 yr of age at time of the index hip fracture; 42% were between 75 and 84 yr of age and one third were >85 yr of age (Tables 1 and 2).
On average, the women were followed for 6.6 ± 6.3 yr (range, 0.0–22.3 yr) with the endpoints of death or still alive at end of the study.
Index hip fracture
The index fractures were distributed as follows: 45% were intertrochanteric (including subtrochanteric), 55% were cervical (intracapsular), and 52% affected the left hip. Women with cervical fractures were significantly younger than those with intertrochanteric fractures (77.7 ± 10.2 yr [range, 31.6–97.4 yr] versus 81.9 ± 9.0 yr [range, 36.2–99.4] yr; p < 0.001). Fractures were treated according to standard procedure at the time; cervical nails, dynamic screw with sliding plate, or intramedullary nailing, whereas only 2% received a total hip replacement. Fourteen (1.8%) women did not undergo surgery (nondisplaced cervical fractures, too frail to survive surgery, or dead before surgery).
At the time of the index fracture, 49 (6.4%) patients had a co-existing fracture: distal forearm (16; 33%), proximal humerus (12; 24%), and other fractures of the upper extremities (7; 14%) or lower extremities (7; 14%).
Prior hip fracture
Previous hip fracture was registered in 84 (11%) women at a mean of 6.3 ± 6.1 yr (range, 0.8–31.4 yr) before the index hip fracture; 60% of these registered within the 5 preceding yr. Women with a prior hip fracture were older than those without a prior hip fracture (82.1 ± 9.0 yr [range, 56.7–96.5 yr] versus 79.3 ± 10.0 yr [range, 31.6–99.4 yr]; p = 0.016).
Of the patients, 717 (94%) were followed from the index hip fracture until death, with only 48 (6%) still alive at the end of follow-up: mean age at death was 86.4 ± 7.9 yr (range, 47.9–104.6 yr) and mean duration of follow-up with death as endpoint was 5.7 ± 5.2 yr (range, 0.0–21.6 yr). This corresponds to an overall survival at 1 yr of 79%, at 2 yr of 71%, and at 5 yr of 48% (Fig. 2A). Women still alive at follow-up were, as expected, younger at time of the index hip fracture than those who died during the study (62.9 ± 9.8 yr [range, 31.6–79.2 yr] versus 80.7 ± 8.8 yr [range, 36.3–99.4 yr]; p < 0.001).
Mortality differed according to age at the time of index fracture. Overall mortality, mortality in 5-yr age bands, and mortality according to age groups are shown in Tables 1 and 2 and Fig. 2B. The 50% mortality in the age strata 75–84 yr was less than one half of that in the youngest age group (<75 yr). Further analysis showed that each added life year at the time of the index hip fracture increased mortality by 8% (p = 0.001). Prior hip fracture did not significantly affect mortality when adjusted for age (OR, 1.05; 95% CI, 0.756–1.20; p = 0.15; Figs. 2C–2F; Table 2).
Mortality was lower in women with cervical fracture compared with those with intertrochanteric index hip fracture (p < 0.001); however, the difference did not remain significant after adjustment for age (p = 0.49).
During follow-up, 342 (45% of 766) women suffered at least one new fracture, making the overall incidence 112 per 1000 person-years after the index hip fracture. The proportion of women with incident fractures was doubled in the youngest age group compared with the oldest (Fig. 1; Tables 2 and 3).
In total, new fractures occurred at 715 occasions (2.1 ± 1.4 occasions/patient; range, 1–10 occasions/patient). The majority of women with incident fractures (71%, 243/342) sustained their fractures at only one or two new fracture occasions, a finding more common in the older age groups, except among patients with prior hip fracture (Table 2).
Fifteen percent of the total number of fracture occasions occurred within the first year after the index hip fracture: 106 fracture occasions resulted in 116 fractures occurring in 94 women (27% of those fracturing). Women with incident fractures were younger at the time of the index fracture compared with those who did not fracture (77.2 ± 9.8 yr [range, 31.6–94.8 yr] versus 81.6 ± 9.6 yr [range, 36.2–99.4 yr]; p < 0.001). This corresponds to a longer mean follow-up for women with incident fractures (9.8 ± 6.3 yr [range, 0.3–22.3 yr] versus 4.1 ± 5.0 yr [range, 0.0–22.0 yr]; p < 0.001).
New fractures and survival:
Among women with incident fractures, 89% (n = 305) died within 8.4 ± 5.2 yr (range, 0.3–21.6 yr). This results in an overall first year survival in the fracture group of 97%, 93% at 2 yr, 72% at 5 yr, and 11% at the end of follow-up. The 50% survival was 8.3 yr. Survival was shorter among women who did not sustain any additional fracture: 1 yr, 64%; 2 yr, 52%; 5 yr, 29%; follow-up, 3%. The 50% survival was 2.2 yr in women without fracture. This can also be compared with the overall mortality rates in Table 1 and Fig. 2A. There was no difference in fracture incidence between women with cervical and intertrochanteric index fractures, despite women with intertrochanteric index fractures being older (p = 0.41, χ2 test).
Total number of new fractures:
As described above, 45% of the patients suffered a subsequent fracture: in total, 768 new fractures were observed after the index fracture (2.3 ± 1.5 fractures/patient; range, 1–11 fractures/patient; Fig. 1; Table 3).
Of all incident fractures, most (55%) occurred within the first 5 yr. However, this was clearly influenced by age, with the highest occurrence in the oldest patients within this relatively short time frame (Table 2).
For the entire cohort, the observed absolute fracture risk was 33% at 5 yr, 40% at 10 yr, and 45% at the end of follow-up (22 yr; Fig. 3A). The risk of new fractures was similar between the three age groups during the first 2 yr; thereafter, the risk remained virtually constant in those >85 yr of age, probably because of mortality. The absolute fracture risk in the age groups <75 and 75–84 yr was similar up to 7 yr (Fig. 3B).
Based on the length of the observation period and on the high mortality, estimates of fracture risk were also adjusted for mortality. Adjusted for mortality, fracture risk in the entire cohort was 45% at 5 yr, 65% at 10 yr, and 86% at the end of follow-up (Fig. 3C). When comparing the three age groups, using Wilcoxon rank test, the mortality-adjusted fracture risk was similar for all ages during the first 4 yr. Using the log-rank test, fracture risk was increased in the 75–84 yr age group compared with those <75 yr (p = 0.012). Hence, the initial mortality-adjusted fracture risk was not explained by age (Fig. 3D).
Absolute 10-yr fracture risk and mortality-adjusted 10-yr fracture risk are shown in Table 2 and reached 70% in those >75 yr of age. In the oldest group, few patients were still alive (at risk), and therefore, these risk estimates are less reliable. Likewise, because of mortality, only 5-yr risk is calculable in those with prior hip fracture; the unadjusted risk is 35%, and the mortality-adjusted 5-yr risk is 53% (data not shown).
Forty-five percent of the patients (n = 341) did not undergo any subsequent skeletal X-ray examination related to trauma or fracture. These women were older (82.1 ± 9.6 yr [range, 36.3–99.4 yr] versus 77.6 ± 9.7 yr [range, 31.6–95.6 yr]; p < 0.001) and had a shorter life expectancy (p < 0.001) compared with those who underwent additional examinations. Among women without additional radiological examinations caused by skeletal trauma, 1-yr survival was 57%, 2-yr survival was 45%, 5-yr survival was 24%, and survival at follow-up was 3%. The 50% survival was 1.6 yr.
Subsequent X-ray examinations were performed in 425 patients (55%) on at least one occasion because of skeletal trauma and suspected fracture. Eighty-three of these patients (20%, 83/425) sustained only a contusion and no fracture (Fig. 1). Radiographic examinations were performed at a total of 1226 occasions (2.9 ± 2.2 occasions/patient; range, 1–14 occasions/patient) and included 1715 regional X-ray examinations (4.0 ± 3.5 X-ray examinations/patient; range, 1–23 X-ray examinations/patient).
This means that in 6 of 10 patients seeking hospital care for musculoskeletal trauma, the X-ray examination verified a fracture, and that at each X-ray occasion, excessive structures were examined in 55% without structural skeletal damage.
Low-energy trauma was the major reason for X-rays after index hip fracture (99%; n = 1217), and 99% of all incident fractures were caused by low-energy trauma (n = 710) regardless of age. Only five (0.7%) of the incident fractures were caused by high-energy trauma, and one (0.1%) was pathological.
In this study of hip fracture, which, to our knowledge, is the first with a remaining lifetime fracture perspective, 45% of all women with a hip fracture sustained at least one subsequent fracture. Fracture risk is dependent on survival both in the short and long term; in the short term, mortality leads to relatively fewer fractures in the very elderly, and in the long term, a younger age at the time of the index fracture leads to an increased fracture risk because of an extended time at risk. This indicates the importance of including age, preferably in combination with a biological age estimate, when developing intervention algorithms.
The initial high mortality rate after hip fracture is well known from other studies.[4-11] However, the long-term mortality rate (i.e., >10 yr or in a lifelong perspective) after a hip fracture has only been briefly described.[21, 24, 31] In this study, we chose to include all adult patients with hip fracture but categorized them into either 5-yr age bands or subdivided them into age groups with the purpose of fully identifying the age dependence on short- and long-term mortality and subsequently on fractures. We can confirm a high initial mortality, predominantly in the oldest patients, where 40% are dead within the first year compared with <3% in those <65 yr of age. This is mirrored in the mean survival rate that is 11, 5, and 2 yr in the respective age groups. Previous hip fracture, on the other hand, only affected immediate mortality in the very old. Within the scope of this study, we described the relative distribution of mortality as a background factor for the evaluation of fracture risk, and as such, we can not draw conclusions regarding mortality directly attributable to the hip fracture or to the influence from other co-existing medical conditions. However, it has been estimated that up to 24% of the deaths during the first year after a hip fracture were causally related to the hip fracture event.
Subsequent fractures and fracture risk
The majority of hip fracture patients are between 75 and 84 yr of age. In absolute numbers, these women suffer the majority of subsequent fractures, whereas the relative risk is higher in those <75 yr of age because of longer exposure time and, in this respect, our findings confirm those of others.[22, 23] Based on the very long observation period in our study, it is clear that the relative risk becomes more pronounced over time and is most evident in younger individuals.
In general, every other woman with a hip fracture suffers a new fracture, but, importantly, only one in three of those are >85 yr of age. Time to fracture has implications in terms of treatment; of all new fractures, one half of all fractures occurred within 4 yr, with one third occurring within 2 yr of the hip fracture. Again, this was influenced by age, and in the oldest, one half of all fractures occurred within 2 yr. Of those fracturing, more than one half suffered more than one fracture, whereas few (8%) suffered more than four subsequent fractures. However, from these data, it is not possible to analyze the different types of subsequent fractures; hence, we can not yet evaluate specific risks, such as if a certain fracture predispose to a similar fracture or to another type of fracture.[23, 37] Because most new fractures occurred among those between 75 and 84 yr of age, an age span with a reasonable expected survival time, a large proportion of incident fractures should be preventable with modern pharmacological interventions in this group. Our findings also indicate that other interventions, with immediate effects and independent of onset time (e.g., falls prevention programs or hip protectors) should be directed toward the very elderly with hip fracture.
Other studies with shorter follow-up have shown that the highest risk of new fractures occurs shortly after a previous osteoporotic fracture, yet, without addressing age-related differences or duration of follow-up, the assumptions are only valid for a limited period of time.[22, 24] In a 10-yr perspective, we found that 90% of all subsequent fractures in the age group of 75–84 yr occurred within this time frame, whereas the equivalent for the age group <75 yr is 65% (data not shown). Residual lifetime risk of fracture has previously been analyzed using estimates from shorter individual follow-ups.[30, 31, 38, 39] Because we followed 94% of the patients over their remaining lifetime, our fracture risk calculations can be transformed into residual lifetime fracture risk. The remaining lifetime fracture risk in the entire cohort was (after 22 yr) 45%, which, after adjustment for mortality, increased to 85%. This means that the chance of a fracture-free survival after hip fracture is 15% if the patient is <75 yr of age at the time of the index fracture and lives beyond 20 yr. Claims are made that 10-yr risk algorithms should be used. Expressed as absolute 10-yr fracture risk, the overall risk is 40% in women with hip fracture and, after adjustment for mortality, 65%. However, with advancing age, a 10-yr estimate becomes less meaningful because of few survivors. Interestingly, regardless of age at the index fracture, the risk of fracture was similar during the first 2 yr after the index hip fracture and diverged later.
In comparison with other studies, our patients were individually followed, whereas others rely on statistical estimates. Because we can only evaluate fracture risk from the index event and not based on the total lifetime fracture risk, comparison with other studies is approximate. A report from Nguyen et al. estimated an overall fracture risk of 58% for women at age 60, a risk that is similar to the 10-yr risk of the youngest age group of our cohort starting with a hip fracture. Another study, based on a fracture audit, showed that patients of younger age had a greater relative risk of subsequent fractures than those of older ages, despite a follow-up time varying from 1 wk to 12 yr. This is highlighted by our findings of a decreasing absolute risk unless adjusting for mortality.
X-ray occasions and trauma without fracture
As expected, most new fractures after the index hip fracture were caused by low-energy trauma. Recognizing the difficulties to estimate risk situations (i.e., falls), we used X-ray occasions as an indirect indicator of musculoskeletal trauma. This is based on the assumption that only musculoskeletal trauma considered severe enough to cause a fracture will be X-rayed; for this reason, the X-ray examination is interpreted as a sign of a fall with a potential fracture risk. However, X-ray occasion might also indicate access to an X-ray facility, the socioeconomic status of the country/patient, and/or the patients'/physicians' anxiety of fracture in someone who has already had a hip fracture. Although to our knowledge, this has not previously been used as a measure, we consider this an interesting indicator not only of trauma but of health care consumption in this population.
In this cohort, the number of X-ray examinations was more than twice the number of fractures, with a fracture diagnosed at 6 of 10 occasions. However, 45% of the women did not undergo any additional X-ray examinations, and they were significantly older and had the shortest postfracture survival, indicating pronounced frailty in this group. Similarly, 20% of those undergoing one or more X-ray examinations but without any new fracture diagnosed were older and had a shorter survival compared with those that later fractured.
Patients with multiple musculoskeletal traumas requiring X-ray examination are at high risk of fracture; 80% did subsequently fracture. Based on the survival data, this group consisted of persons either initially relatively healthy or frail persons sustaining both trauma and a second fracture soon after the index fracture. Previous reports described fallers, as older, with other pathologies, and with higher degree of dependency, but used a limited observation period and are likely to not include those who are healthier in relative terms.[32, 33, 35]
The patients in this study sustained the index fracture before the availability of specific antiresorptive osteoporosis treatment; hence, we report on the natural risk pattern after hip fracture. However, identification and initiation of pharmacological osteoporosis treatment after fracture is very low, and the effect on fracture reduction at the population level is unknown.[17, 40] The incidence of prior hip fracture and the rate of co-existing fractures were similar to that of others.[41-43] Despite being older, and contrary to our expectations, mortality in those with prior hip fracture was similar up to 5 yr after the index fracture. In this group, however, fracture risk was higher, as was the proportion with multiple fractures. In a recent Finnish study, the age distribution was similar; in addition, factors potentially contributing to a second fracture such as worse walking ability and more medications affecting gait were described, which are factors that we are unable to ascertain.
Strengths and limitations
The strengths of this study include the exceptionally long follow-up, the population-based inclusion, and the homogeneity of the population. We also were able to identify virtually all radiographic examinations that each person has undergone during their remaining lifetime. Our study is comprehensive, and the data were obtained directly rather than relying on databases or mail surveys. This means that we are uniquely able to evaluate the true rather than estimated fracture incidence and fracture risk after hip fracture. All data were verified by one person, ensuring consistency in reporting and interpretation of fracture and X-ray information is high. The size of the cohort is large enough to allow reporting age-based subgroups and actual years of follow-up. In addition, the Swedish personal identification number allows for obtaining the mortality date of all individuals which may be difficult in other settings. Subsequently, it is our current understanding that no other study has followed this number of patients to the end of their life.
Our study has some limitations; one might argue that, whereas this study has a long observation period, changes in the management of hip fracture patients with respect to surgical practices, anesthetics, and postoperative management, including treatment of osteoporosis, makes the results less relevant. To an extent this may be true; however, it should merely slightly shift the survival curves, equally for all age groups, and hence not influence the overall conclusions. Furthermore, life expectancy in women has increased by 2 yr over the past 20 yr, and an increased life expectancy should subject them to a longer time at risk. This has not been explored; however, the mean age of suffering a hip fracture in Sweden has equally increased by 2 yr. The aim of this study was to evaluate survival and fracture risk at different ages in a hip fracture population. It would have been advantageous to also include a control group to evaluate excess mortality, but this was not within the scope of this study.
In this comprehensive study of fracture risk and mortality after hip fracture in women, we followed almost 95% of the patients until death; hence, we are observing the true fracture incidence and long-term survival for the majority of patients.
Age and survival are the most important factors for future fracture risk in women with hip fracture, and our study emphasized that age stratification and survival need to be taken into account when developing fracture prevention programs and that treatment should be initiated rapidly, particularly in those between 75 and 85 yr of age or with a corresponding biological age.
Support for the study was received from the Swedish Research Council (Grant K2006-73X-14691-04-3), Greta and Johan Kock Foundation, A Påhlsson Foundation, A Osterlund Foundation, Malmö University Hospital Research Foundation, Research and Development Council of Region Skåne, Sweden, and the Swedish Medical Society. We thank Jan-Åke Nilsson for statistical advice.