Osteoporosis and the resulting fragility fractures are a substantial and increasing source of morbidity in Western populations. Fractures of the distal part of the radius are common, and their incidence, particularly in females, increases with age1. The role of osteoporosis and osteopenia in the pathogenesis of fractures of the distal part of the radius is well established1-3. We were unable to find previous clinical studies investigating the link between osteoporosis and fracture severity or whether low bone mineral density is related to radiographic outcomes after injury.
Although there is controversy surrounding the optimal surgical treatment for displaced fractures of the distal part of the radius, there is general agreement that, to maximize functional recovery, the primary goal of treatment should be to regain as close to an anatomical position as possible4-7. More severe fractures can be considered to be those that lead to the highest likelihood of malunion developing or that require surgery to prevent malunion. Algorithms that allow one to calculate the percentage probability of early and late fracture displacement, late adaptive carpal malalignment, and malunion on the basis of clinical variables and radiographic measurements have recently been published8. For the purpose of this study, the percentage probability of any given fracture being followed by one of these adverse outcomes served as a surrogate measure of fracture severity.
The aim of this study was to investigate the association between the degree of osteoporosis and the severity of distal radial fracture.
A prospective study of a consecutive series of patients with a low-energy distal radial fracture (defined as a distal radial fracture sustained in a fall from a standing, or lower, height) who underwent dual x-ray absorptiometry scanning of the proximal part of the femur was performed. All patients who sustain a fracture in the geographic region served by our hospital are treated in a single orthopaedic trauma unit. All patients over the age of fifty-five years with a fracture of the distal part of the radius were offered dual x-ray absorptiometry scanning of the hip, and those who consented to undergo the scanning were selected for this study. Dual x-ray absorptiometry scans were performed three months after the injury, and, at that stage, the radiographs from the time of the injury were retrieved for analysis.
The T-score for the left hip was used as the standard, but in ten cases the left hip had been replaced or was not suitable for study so the measurement was made from the right hip. The T-score is defined as the number of standard deviations by which the recorded bone density differs from a control value derived from the mean and standard deviation of values in a healthy young adult population9. In this study, we used the World Health Organization's definitions of osteoporosis (a T-score of less than -2.5), osteopenia (a T-score of -1 to -2.5), and normal bone density (a T-score of more than -1).
From August 2005 to May 2006, 158 consecutive patients with a distal radial fracture underwent dual x-ray absorptiometry scanning. Anteroposterior and lateral radiographs of the wrist made at the time of presentation, after manipulation in the Emergency Department, and at one, two, and six weeks were required for a patient to be included in this study. This complete series of radiographs was unavailable for twenty-one of the 158 patients. The remaining 137 patients are the subjects of this study.
The radiographs were assessed manually with use of a protractor and ruler. Dorsal angulation10, ulnar variance11, and carpal malalignment4 were calculated. Metaphyseal comminution was a qualitative measurement that was recorded as involving the dorsal aspect of the metaphysis, the volar aspect of the metaphysis, or both. Intra-articular displacement was considered to be present if there was intra-articular extension of the fracture creating a step-off of =2 mm in the articular surface. Carpal malalignment was defined as the failure of the long axes of the capitate and radius to intersect within the carpus as seen on the lateral radiograph of the wrist. These measurements are used in standard clinical practice for the assessment of distal radial fractures in our unit, but to minimize the risk of interobserver variability a small pilot study was carried out before the start of this investigation. In the pilot study, the first author (R.A.E.C.) performed measurements on ten sets of radiographs, which were then checked by the senior author (M.M.McQ.). No interobserver differences were found in the measurements of the categorical variables, the interobserver variation in the measurements of angulation was never >3°, and the interobserver variation in the measurements of ulnar variance was never >1 mm. The first author then performed all of the radiographic measurements with the exception of the AO/OTA classification. The senior author also checked a random sample of 10% of the measurements during the study, and the interobserver variation was the same as that in the pilot study. Fractures were classified by group and subgroup according to the AO/OTA system12 solely by the senior author.
Displacement was defined as >10° of dorsal angulation or >3 mm of ulnar positive variance on the initial radiographs. An acceptable reduction was defined as dorsal angulation of =10° and ulnar positive variance of =3 mm after closed manipulation. Redisplacement was defined as a loss of position into dorsal angulation of >10° or ulnar positive variance of >3 mm after a previously successful closed manipulation into an acceptable reduction. Early instability was defined as redisplacement occurring within two weeks after the injury and initial closed reduction. Late carpal malalignment was defined as the presence of carpal malalignment at six weeks or at the time of operative intervention. Malunion was defined as displacement at six weeks when a fracture had not demonstrated early instability. It was also defined as the theoretical risk of a fracture being displaced at the time of union if the fracture was untreated other than by closed manipulation at presentation. Therefore, for the purpose of quantifying fracture outcomes in this study, malunion included two groups of fractures. The first group consisted of fractures that were manipulated but redisplaced so that they were in an unacceptable position at six weeks. The second group consisted of fractures that were manipulated and redisplaced but underwent surgery because an unacceptable position would have occurred at union if no further intervention had taken place.
The severity of the fractures was quantified with use of the published algorithms for calculating the probability of early and late displacement, late carpal malalignment, and malunion8. Our definitions of early instability, carpal malalignment, and malunion were the same as those used in the reported algorithms. These formulae were developed on the basis of multivariate analysis of a range of clinical and radiographic variables measured at the time of presentation of distal radial fractures and were used to generate a percentage probability that the fracture in question will lead to early instability, malunion, or carpal malalignment. We used these percentages as our quantitative measures of fracture severity. The input variables for these formulae are patient age, ability to live independently, AO/OTA subgroup, presence of dorsal or volar comminution, dorsal angulation of the distal radial fragment, and ulnar variance at presentation. Hence, these were the data collected for this study.
Undisplaced fractures were immobilized initially in a plaster-of-Paris dorsal slab. Fractures with volar angulation or an articular step-off of >2 mm underwent internal fixation at the discretion of the treating surgeon. Displaced fractures underwent closed reduction with the use of intravenous regional anesthesia. The fractures were subsequently reassessed radiographically at one week and two weeks after injury. If there was no loss of position at the one-week review, the forearm dorsal slab was converted to a short arm cast. This type of treatment regimen has been standard practice in our institution and in many other parts of the United Kingdom for many years. Fractures with unacceptable sagittal or coronal angulation, radial shortening, or intra-articular displacement at the one-week or two-week review underwent surgery at the discretion of the treating surgeon.
The dual x-ray absorptiometry T-scores for all patients were collected in a prospective database by a dedicated research nurse. At this stage, the details regarding the patients in the study were separated from the results of the dual x-ray absorptiometry scanning to ensure that the subsequent data collection and analysis could be performed with the authors blinded to the T-scores. The clinical notes on the patients identified for participation in the study were retrospectively reviewed to ensure the exclusion of patients with high-energy injuries and to confirm which patients had undergone surgery.
Statistical analysis was performed with use of a proprietary statistical software package. The dual x-ray absorptiometry T-scores were analyzed with use of the Shapiro-Wilk test, which showed that the data were not normally distributed; therefore, nonparametric tests were used. The T-scores were correlated with the fracture severity scores (the probability of early instability, late carpal malalignment, or malunion) with use of the Spearman rank-order correlation test and the Pearson product-moment correlation. Multivariate regression analysis was also performed. Comparisons among groups of patients categorized by T-score were performed with use of the Mann-Whitney U test. A p value of <0.05 was considered to be significant.
Source of Funding
There was no external source of funding for any of the work in this study.
One hundred and twenty-seven (93%) of the 137 patients were women. The mean age was seventy-one years (range, forty-five to eighty-seven years). With use of the AO/OTA system, sixty-four fractures were graded as type A, seventeen were graded as type B, and fifty-six were graded as type C12. There were no open fractures.
Observed Rates of Early Instability, Carpal Malalignment, and Malunion Compared with Predicted Rates
Forty-five (33%) of the 137 fractures had early instability and underwent surgery. Twenty-five underwent nonbridging external fixation, ten underwent volar plate fixation, eight underwent bridging external fixation with Kirschner wire augmentation, and two underwent bridging external fixation without Kirschner wires. Another thirty-seven fractures were displaced at or after the six-week review; thus, a total of eighty-two (60%) of the 137 fractures were defined as malunited for the purposes of this study. Fifty-three fractures (39%) were associated with adaptive carpal malalignment at the time of surgery or at six weeks. The mean percentage probability of early instability according to the calculations with use of the published algorithms8 for the 137 fractures in this series was 35%, the mean percentage probability of malunion was 56%, and the mean percentage probability of carpal malalignment was 31%.
Correlation Between Dual X-Ray Absorptiometry T-Scores and Predicted Early Instability
Univariate analysis with use of the Spearman rank-order correlation test showed a significant correlation between an increasing degree of osteoporosis (increasingly negative femoral neck T-score) and the calculated probability of early instability. The rank correlation coefficient (R) was —0.24 (p < 0.01). The line of best fit according to the Pearson product-moment correlation gives the formula: P = 21.2 — (T × 6.4), where T is the T-score and P is the percentage probability of early instability. A fracture in a patient with a T-score of -1 (borderline osteopenia) would have a 28% chance of having early instability. This increases to 37% for a T-score of —2.5. When a multivariate analysis was used to correct for the effect of age, the same correlation was seen and remained significant (beta = -0.19, p = 0.033). The T-scores were also categorized as representing normal bone mineral density, osteopenia, or osteoporosis. Figure 1 shows the probability of early instability for the three categorical groups (normal bone mineral density, osteopenia, and osteoporosis). The mean probability of early instability was 43% in the osteoporosis group, 35% in the osteopenia group, and 28% in the normal group. The Kruskal-Wallis analysis of variance by ranks test for nonparametric data showed a significant difference in the probability of early instability among the three groups (p = 0.024). A Mann-Whitney U test comparing the individual groups showed a significant difference in the probability of early instability between the patients with osteoporosis and those with normal bone mineral density (p = 0.012). The difference between the normal group and the osteopenia group approached significance (p = 0.056), and the difference between the osteopenia group and the osteoporosis group did not reach significance (p = 0.18). A post hoc power analysis with use of a one-way analysis of variance power calculation showed a power of 0.82, suggesting that the study was sufficiently powered to detect a difference between these values.
Correlation Between Dual X-Ray Absorptiometry T-Scores and Predicted Late Carpal Malalignment
Univariate analysis with use of the Spearman rank-order correlation test showed a significant correlation between an increasing degree of osteoporosis (increasingly negative femoral neck T-score) and the calculated probability of carpal malalignment. The rank correlation coefficient (R) was —0.32 (p = 0.00015). The line of best fit according to the Pearson product-moment correlation gives the formula P = 24.1 — (T × 4.5), where T is the T-score and P is the percentage probability of carpal malalignment. A fracture in a patient with a T-score of -1 (borderline osteopenia) would have a 29% chance of leading to carpal malalignment. This increases to 35% for a T-score of -2.5. When a multivariate analysis was used to correct for the effect of age, the same correlation was seen and remained significant (beta = —0.20, p = 0.020). Figure 2 shows the probability of carpal malalignment for the three categorical groups (normal bone mineral density, osteopenia, and osteoporosis). The mean probability of late carpal malalignment was 39% in the osteoporosis group, 31% in the osteopenia group, and 25% in the normal group. The Kruskal-Wallis analysis of variance by ranks test for nonparametric data showed a significant difference in the probability of carpal malalignment among the three groups (p = 0.006). A Mann-Whitney U test comparing the individual groups showed significant differences in the probability of carpal malalignment among all three groups (p = 0.00025 for the difference between the patients with osteoporosis and those with normal bone mineral density, p = 0.028 for the difference between the normal group and the osteopenia group, and p = 0.013 for the difference between the osteopenia group and the osteoporosis group).
Correlation Between Dual X-Ray Absorptiometry T-Scores and Predicted Malunion
Univariate analysis with use of the Spearman rank-order correlation test showed a significant correlation between an increasing degree of osteoporosis (increasingly negative femoral neck T-score) and the calculated probability of malunion. The rank correlation coefficient (R) was -0.29 (p = 0.00079). The line of best fit according to the Pearson product-moment correlation gives the formula P = 41.0 — (T × 7.3), where T is the T-score and P is the percentage probability of malunion. A fracture in a patient with a T-score of -1 (borderline osteopenia) would have a 48% chance of malunion. This increases to 59% for a T-score of -2.5. When a multivariate analysis was used to correct for the effect of age, the same correlation was seen and remained significant (beta = -0.18, p = 0.042). Figure 3 shows the probability of malunion for the three categorical groups (normal bone mineral density, osteopenia, and osteoporosis). The mean probability of early instability was 66% in the osteoporosis group, 56% in the osteopenia group, and 48% in the normal group. The Kruskal-Wallis analysis of variance by ranks test for nonparametric data showed a significant difference in the probability of carpal malalignment among the three groups (p = 0.019). A Mann-Whitney U test comparing the individual groups showed significant differences in the probability of carpal malalignment among all three groups (p = 0.00073 for the difference between the patients with osteoporosis and those with normal bone mineral density, p = 0.036 for the difference between the normal group and the osteopenia group, and p = 0.033 for the difference between the osteopenia group and the osteoporosis group).
Correlation Between Dual X-Ray Absorptiometry T-Scores and Observed Malunion and Early Instability
Analysis with use of the Spearman rank-order correlation test showed a significant correlation between an increasingly negative femoral neck T-score and the observed occurrence of malunion. The rank correlation coefficient (R) was -0.18 (p = 0.042). When the same analysis was performed with carpal malalignment as the end point, the effect size was found to be small and to not reach significance (R = -0.13, p = 0.14). Similarly, with early instability as the end point, the effect size did not reach significance (R = -0.08, p = 0.35).
Correlation Between Dual X-Ray Absorptiometry T-Scores and AO/OTA Fracture Type
The mean T-score was -1.75 ± 0.98 for the patients with an AO/OTA type-A fracture, -1.36 ± 1.24 for those with a type-B fracture, and —1.20 ± 0.93 for those with a type-C fracture. This trend for increasing (less negative) bone mineral density with the type-B and C fractures reached significance (R = -0.26, p = 0.0026). No significant correlation between T-scores and the nine AO/OTA groups or subgroups was found.
It is well established that reduced bone mineral density is associated with an increased incidence of many types of fractures. A fracture of the distal part of the radius is one of the most common fragility fractures and is definitely associated with osteoporosis1-3.
Low-energy distal radial fractures most commonly result from a fall onto an outstretched wrist. Loss of bone mineral density in the distal part of the radius leads to a reduced ability of the bone to resist these deforming forces and hence to the increase in fracture incidence previously observed at this site. It therefore follows that the same loss of ability to resist deforming forces would lead to increased comminution, displacement, and angulation at the fracture site. This theory was confirmed in a cadaver study by Lill et al., performed under standardized laboratory conditions, in which a correlation was observed between decreasing bone mineral density and increasing fracture severity13. It is the early or late loss of position of the fracture that leads to the clinically important end point of malunion or the requirement for surgery.
Itoh et al.14 found an association between decreasing bone mineral density and decreased radial length at union, but they found no association with any other features of fracture severity such as the Frykman classification or any clinically important end points such as a need for remanipulation or late surgery for the treatment of malunion. However, they excluded from their study all patients with injuries not amenable to closed reduction and cast treatment, which would have been defined as being among the more severe injuries in our series.
The results of our study show associations between loss of bone mineral density and fracture severity, with most reaching significance in the comparisons that we performed. Furthermore, these associations remained even when we corrected for patient age. However, the effect sizes (the R values) were low, suggesting that osteopenia is not the only variable contributing to fracture severity. In the correlation tests that we used, the R values range from 1 for a 100% perfect positive correlation to 0 for no correlation and -1 for a 100% negative correlation. Our analyses showed significant negative R values, confirming that the association between osteopenia and fracture severity is real, but the most strongly negative score was only -0.36. This is presumably because there are factors in addition to reduced bone mineral density that contribute to the severity of a fracture but cannot be measured in the normal clinical setting and therefore cannot be accounted for in a multivariate analysis. The first of these factors is the energy of the injury. All patients in this study had a low-energy injury, defined as a fall from standing, or lower, height. However, within this definition, there is still a wide range of energy transmitted. Another factor determining the severity of an injury is the position of the wrist and the angle of the forearm at impact. In particular, these determine the type of angulation and metaphyseal comminution that commonly determine stability and hence the requirement for surgery. It is interesting to note that, in the cadaver study by Lill et al., in which the investigations were performed in a standardized laboratory setting, the effect sizes ranged from 0.09 to 0.7013. The relatively low effect sizes seen in our study do not diminish the potential clinical relevance of our findings.
The estimates of the probabilities of early instability, carpal malalignment, and malunion used in this study were calculated with published formulae that had been derived from analyses of approximately 4000 fractures8. Mackenney et al. performed a multiple logistic regression analysis of clinical and radiographic features of the injury at presentation in order to calculate formulae predicting the probability that an individual fracture would be followed by early instability, late carpal malalignment, or malunion8. It is important to remember that the definition of malunion used in this study included not only patients with unacceptable alignment at union but also those who underwent surgery because it was assumed that they would have an unacceptable position at union if no surgery took place. Metaphyseal comminution, ulnar positive variance, and dorsal angulation were the most significant predictors of these complications. The calculated probabilities of early instability, late carpal malalignment, and malunion therefore represent a quantitative measure of a weighted combination of these factors. The similarity between the rates of early instability, carpal malalignment, and malunion predicted from the formulae and the observed rates helps to validate these formulae.
Use of these formulae to quantify fracture severity provides results as continuous variables (percentages), which allows a more powerful statistical analysis than is possible with the use of discrete variables. The observed occurrence of early instability or malunion is an all-or-nothing discrete variable (the end point was either present or absent). This study was probably underpowered to detect a significant difference in the observed rate of early instability, but a significant difference in the rate of malunion was observed with decreasing bone mineral density. It is not possible to comment on the observed rate of late carpal instability because the surgery that was performed altered the natural history of the fracture.
Completely articular fractures (AO/OTA type C) are generally considered to be more severe than extra-articular fractures (type A). Our finding that the bone mineral density in patients with a type-C fracture was higher than that in patients with a type-A fracture might therefore at first sight appear counterintuitive in the light of our other findings. There are two possible reasons for our finding. First, the fracture types that are most severe according to the definition of the formulae used in this study are those with impaction (hence increased ulnar positive variance) or metaphyseal comminution, as these factors lead to the greatest increase in the risk of instability, malunion, and carpal malalignment. These fractures are classified as AO/OTA types A3, C2, and C3. Second, it may be that, in patients with osteoporosis, bone mineral density is reduced more in metaphyseal bone than it is in subchondral bone, leading to a tendency toward metaphyseal (type-A) fractures in preference to intra-articular (type-B and C) fractures. It is therefore perhaps not surprising that we found the bone mineral density in the patients with an extra-articular fracture to be lower than that in the patients with an intra-articular fracture.
There are three possible reasons for failure to detect a correlation between decreasing bone mineral density and the nine AO/OTA fracture groups. First, even with experienced observers there is known to be variability in the classification of groups and subgroups with use of the AO/OTA system15,16. Second, the study is underpowered to detect differences among the number of groups. Third, it is difficult to rank all AO/OTA groups in order of severity. For example, although it is evident that a fracture with extensive articular comminution (type C3) should be considered more severe than a simple radial metaphyseal fracture (type A2), it is not clear whether a simple, complete articular fracture with no displacement (C1) should be considered more severe than an extra-articular fracture with extensive metaphyseal comminution (A3).
The end points used in this study are clinically relevant. Early instability is usually considered to be an indication for surgery, although there are other factors dictating whether surgery is required. Malunion is generally accepted to be associated with poorer functional outcomes5,6, and carpal malalignment is associated with pain, loss of range of motion, and reduced grip strength4,17. However, it is not clear whether these findings can be generalized to other osteoporotic fractures, such as those in the proximal part of the humerus, the proximal part of the femur, and the ankle.
This study has some limitations. Our quantification of fracture severity was based on probabilities derived from algorithms based on radiographic measurements, without use of clinical outcome data such as function scores or scores on overall health questionnaires. We are not aware of any published data validating these algorithms in a clinical setting. Therefore, we can comment only on an association between reduced bone mineral density and radiographic fracture severity, not clinical fracture severity. Another limitation is that most of the radiographic measurements were performed by a single person. All of us are very familiar with these measurements, as they are used as a standard part of the clinical practice in our unit, and our pilot study suggested that interobserver variability was low. This was confirmed by the sample, consisting of 10% of the radiographs, that was double-checked by the senior author. Nonetheless, a risk of interobserver or intraobserver error in the measurements remains. As far as we are aware, these are the first data to show an association between bone mineral density and the severity of any fracture type. 