Fractures were initially reduced and sugar-tong splint immobilized in the emergency department by an orthopaedic resident. Participants either had an attempt at closed reduction and casting at presentation, or they underwent immediate surgery. Those who had closed reduction and casting were further categorized based on whether or not they underwent surgery within four weeks following reduction. In all cases, the decision to cast and the decision for surgical intervention were the opinion of the attending physician at the time of treatment. Imaging data were obtained from PACS digital radiographs.

The proximal end of the radial shaft was defined as the proximal extent of the bicipital tuberosity. The distal end of the shaft was defined radiographically by measuring the width of the distal radial physis, then plotting that distance proximally on the radius. This is an approximation of the proximal extent of Lister’s tubercle and accounts for age and bone length. Fractures were classified as proximal, middle or distal based on an equal one-third division of the length of the shaft. Ulnar fractures were classified according to the corresponding level of the radius. Angulation of the radius and ulna were measured on AP and lateral radiographs at presentation, post-reduction, and weekly for up to four weeks of follow-up. All measurements were made twice by the same trained research assistant with the mean reported. For participants who underwent surgery or additional fracture manipulation following closed reduction and casting, the radiograph prior to intervention was used for follow-up. Angulation was defined as the total degree of deviation of the distal fragment in relation to the proximal fragment. FIGURE 1 illustrates these measurements.

The primary outcome was the degree of residual angulation present at defined time periods during follow-up. Based on reasonable evidence presented in the literature from both clinical and cadaveric studies, the threshold for acceptable shaft angles for females 8 years or younger and males 10 years or younger were 20 degrees for the distal-third, 15 degrees for the middle-third, and 10 degrees for the proximal-third^{9,13,19,22–26}. In order to account for skeletal maturity according to age and sex, angulations up to 10 degrees were considered acceptable at all shaft levels for females older than 8 and males older than 10 years^{4,7,13,19,22,25}. Failure of non-operative treatment was defined as exceeding these criteria. We also recorded fracture characteristics and the week in which initial radiographic evidence of failure was identified.

To evaluate the mean natural angulation of each segment, as well as the reliability of our measurement methods, we assessed a convenience sample of 30 forearm radiographs considered to represent the patient cohort based on age and sex. The radius was divided into thirds, and angular measurements were created for each segment using the midpoint of the segment as an imaginary fracture line. The principal investigator and trained research assistant independently measured each forearm twice; the second reading occurred one week following the initial reading. Our segmental analysis of the natural bow of the radius revealed an apex radial bow of 1.5 degrees in the proximal-third, 6.0 degrees in the middle-third, and 1.7 degrees in the distal-third. Based on these measurements, we applied a population-based correction factor of 6 degrees apex radial to AP measurements of the middle-third of the radius (i.e. apex radial measurements were decreased by 6 degrees, and apex ulnar measurements were increased by 6 degrees). Corrections were not applied to the distal and proximal segment measurements because their mean values were within the accepted margin of reader error of ±5 degrees^27,28. Inter- and intra-rater reliability was assessed using limits of agreement.

Participants are described in TABLE 1. The mean age was 8.5 years and 63% were male. Bayonet apposition with shortening was present in 27% of cases. For the radius, proximal-third fractures accounted for 24% of fractures, 37% were middle-third, and 39% were distal-third. For the ulna, 16% were proximal-third, 44% were middle-third, and 40% were distal-third fractures. The mean largest angle present on AP or lateral radiograph at presentation was 31 degrees for both the radius and the ulna. The majority of radius factures were apex volar (89%) and apex radial (54%).

In our study population, 144 (51.1%, CI₉₅ 45.1% to 57.0%) of the 282 participants exceeded the acceptable angulation criteria at follow-up. There were 53 patients from the non-operative treatment cohort who later underwent surgery. Of these, one patient did not have angulation data available prior to surgery, 40 had exceeded the angulation criteria, and 12 did not exceed the angulation criteria at the time of surgery (TABLE 2). The mean time to surgery was 10.6 days (SD= 6.7 days). Six patients received fracture re-manipulation with closed reduction and casting.

In the univariate logistic regression analysis, there was a non-linear effect of age with an apparent threshold at 10 years for both males and females; older participants (≥10 years) were more likely to fail (TABLE 3). Those with proximal level fractures, those with a smaller angle at presentation, and those with radius fractures having apex ulnar directionality were also more likely to fail. The odds of failure were nearly three-fold higher for the middle-third radius fractures compared to distal-third fractures (OR= 2.52, CI₉₅ 1.44 to 4.41), and for middle-third ulna fractures compared with distal-third ulna fractures (OR= 2.88, CI₉₅ 1.69 to 4.92)

The multivariate model (inclusive of all main effects) suggested that age, level of fracture for the radius, and angle of fracture for the ulna were the key predictors of failure (TABLE 4). The overall C-statistic for this model was high (0.814, CI₉₅ 0.763 to 0.864). Without considering complex interactions, the final predictive model included three binary variables: age ≥10 years v. 0–9 years, angle of fracture for the ulna <15 degrees v. ≥15 degrees, and level of fracture for the radius proximal v. middle and distal (Table 5). The C-statistic for this derived model was 0.762 (CI₉₅ 0.705 to 0.818).

Of the 144 participants who failed closed reduction and casting within the four week period, 80 (55.6%) had their first radiographic evidence of failure during the first week post-reduction, 34 (23.6%) failed during the second week, 23 (16.0%) failed during the third week, and 7 (4.9%) failed during the fourth week (FIGURE 2). Immediately following closed reduction and casting, 75 participants (27%) exceeded the angulation criteria, indicating these participants were inadequately reduced. Of these 75, 63 failed to meet angulation criteria at the fourth week of follow-up (84%).

Evaluation of 282 complete both-bone forearm shaft fractures treated by closed reduction and casting revealed that 51% of participants failed to meet angulation criteria with non-operative treatment within four weeks of follow-up. Our findings are consistent with previous studies. Kay et al. determined that 64% of their patients older than age 10 failed closed reduction and casting using an angulation criterion of 10 degrees⁴. Thomas et al. found that 39% of their patients with mid-shaft fractures had unsatisfactory results due to an angulation of >10 degrees, loss of range of motion, or a cosmetic deformity³. The results of our multivariate analysis indicate that those at highest risk of failure following closed reduction and casting are patients 10 years or older (OR= 2.79, CI₉₅ 1.47 to 5.29), those with proximal radius fractures (OR= 6.81, CI₉₅ 3.28 to 14.14), and those with initial ulna angulations <15 degrees (OR= 2.94, CI₉₅ 1.49 to 5.83). While the findings for age and level of fracture are intuitive, it is initially perplexing that a lesser initial angle of the ulna predicts failure, though the association was strong and consistent. We suggest that this is due to the relatively intact ulna allowing uneven static and dynamic forces to negatively influence the position of the radius. Re-angulation occurs as swelling decreases and the tensions afforded by the interosseous membrane and attaching muscles change. The location of attachment and relative strength of these structures will affect the position of the healing bones²².

Over half of participants who failed angulation criteria at follow-up had their first radiographic evidence of failure during the first week, with 95% failing within three weeks. Voto et al. made a similar observation, stating that the majority of fractures that re-angulated in their study did so by two weeks⁸. Our results also showed that 84% of participants who had post-reduction angulations exceeding the acceptable angulation criteria failed to meet the criteria at follow-up, illustrating that residual angulation generally does not improve within the first month following reduction. This suggests that anatomic reduction is imperative. Patients should also be monitored weekly within the first month in order to properly monitor closed fracture care. We note that bayonet apposition with shortening, leading to severe compromise of the interosseous space, was present in all 12 of our participants who underwent surgery though they did not fail the angulation criteria. This shows that clinicians base decisions on factors other than angulation alone. Based on limits of agreement, we found the maximum deviation between any two measurements was 4.3 degrees, which is consistent with previous estimates of variation in radiographic fracture angulation measurements of ±5 degrees^27,28.

The anatomy of the forearm dictates its function. In 1959, Sage described the bow of the radius to be 9.3 degrees³¹. His approximation used the entire bone to estimate the bow of the radius, which averaged over the subtle curvatures of each segment. Clinically however, forearm shaft fractures are analyzed by thirds. After we evaluated 30 normal forearms and found a mean angulation of 6.0 degrees apex radial in the middle-third, we felt it appropriate to account for the natural bow of the radius using a population based correction factor. Since the mean angulation value for the distal and proximal thirds were within the range of error for our measurements, we chose to not correct these segments. In 1986, Roberts employed a similar technique and found a mean angulation of 3.7 degrees in the middle third of the radius³². Other methods of approximating the natural bow of the radius using axis deviation have been described and applied, although the calculations can be cumbersome in practice⁹. Our hope is that future studies of forearm shaft fractures will apply methods similar to ours to account for the anatomy of the radius.

The importance of the normal curvature of the radius is recognized for its role in the normal rotation of the radius about the ulna^6,33–35. Normal range of forearm motion is approximately 71 degrees pronation and 84 degrees supination³⁶. Schemitsch and Richards, who examined both-bone fracture malunions in adults, claimed that loss of forearm rotation and grip strength could be expected if the normal radial bow magnitude and location were not restored relative to the non-injured extremity within 4–5%³⁵. Clinical studies of both-bone forearm shaft fractures in children show that range of motion is significantly affected by residual angulation of fractures. Daruwalla et al. found that 53% of patients had limited range of motion, and 13% had greater than 40 degrees of loss¹⁰. Carey et al. revealed that 60% of their patients 11–15 years had residual loss of motion, up to 30 degrees¹¹. Cadaveric studies described the loss of rotation expected with varying degrees of angulation. Matthews et al. found that approximately 30%, or 60 degrees, loss of pronation-supination resulted with an angulation of 20 degrees⁵. Tarr et al. described angulations of 15 degrees causing greater than 27% loss in range of motion, with middle-third deformities causing the greatest loss⁶. Sarmiento et al. confirmed these findings and showed greater supination losses in middle-third forearm shaft fractures and greater pronation losses in distal-third forearm shaft fractures⁷.

The criteria for failure that we used changed based on age and level of fracture, an approach based on biological phenomena and well supported in the literature^{9,13,19,22–26}. Though paradoxical outcomes occur (poor alignment with no loss of motion and anatomic alignment with significant loss of motion), there is a direct relationship between increasing radial shaft angulation and loss of forearm rotation^5–7. However, the remarkable remodeling potential that the radius and ulna posses in children is an important confounding factor of this relationship²⁶. Clinicians account for remodeling in their treatment decisions, allowing post-reduction angulations of greater than 10 degrees in younger children. Most clinicians agree that prognosis changes as the fracture level progresses from distal to proximal due to a greater capacity for remodeling at the distal end^4,15,32,37. The distal forearm in younger children is particularly adept at correcting non-anatomic alignments because 75% of the radial growth occurs at the distal physis¹⁶. Remodeling capacity declines with age, and is greatly diminished as females reach 8 years and males 10 years of age^16,26. Since remodeling capacity decreases with age and more proximal level shaft fractures, residual angulation of middle and proximal shaft fractures is more problematic in older children^2,15,26. The angulation criteria used in our study account for likely remodeling capacity.

One may argue as to what amount of loss of motion is clinically relevant. Morrey et al. found that most activities of daily living could be performed with only 100 degrees of rotation, 50 degrees of both pronation and supination³⁶. Others have found that patients are often unaware of their deficits of 35 up to 60 degrees loss of forearm motion due to compensation with shoulder abduction, and that cosmesis is often the patient’s main concern^2,10–13. However, even if a child is unaware of a moderate loss of forearm rotation or daily activities are not impaired, it seems inappropriate to place a child at this disadvantage when better options exist. Certainly, informed decision makers would choose normal range-of-motion.

The imperfect correlation between residual angulation and loss of forearm rotation fuels the debate over the proper management of both-bone forearm fractures, particularly in older pediatric populations^11,13. Several studies show non-operative treatment to have low rates of re-angulation^8,37. However, high numbers of distal forearm fractures near the metaphysis skew these results due to their more forgiving nature compared to mid-shaft fractures³⁸. Fracture re-manipulation with closed reduction and casting is another viable option if initial reduction is insufficient. Operative techniques, including ESIN, have been shown to effectively prevent re-angulation^17–21. However, surgery carries its own risk of complications, and it is wise to remember Sarmiento’s advice that closed fracture treatment must not be abandoned for newer, less proven, surgical techniques, and that surgical intervention does not always restore a normal range of forearm motion^7,39. The current treatment protocol used at our institution implements the same criteria for acceptable shaft angles as indicated in our study and mimics the criteria proposed by Price et al.¹³. Bayonet apposition is considered acceptable, unless the interosseous space is significantly compromised, and malrotation of less than 45 degrees is accepted. Our audit of surgical decision making at our institution revealed that only 40 of the 144 participants who did not meet radiographic angulation standards actually underwent surgery. Barriers to demanding optimal radiographic outcomes include limited operating room access, patient choice, lack of adequate follow-up, perceived potential for remodeling, lack of noticeable range of motion loss despite angulation, and differing opinions among surgeons in our practice.

Our work must be interpreted within the context of the study design. Our predictive model was derived from a retrospective cohort, and future prospective validation is required. Another limitation of our study is that we used angular data from only two planes. Concurrent angulation in both the radioulnar and dorsoventral planes indicates a maximum angle that is out of the plane of either view, whereby the actual magnitude and direction can be approximated using geometry⁹. We also note that a cohort of our both-bone shaft fractures went immediately to surgery, suggesting that our prediction of failure may in fact be underestimated due to removal of the potentially more severe cases. We acknowledge that re-angulation may occur beyond one month of follow-up, though our results indicate it is more likely early in treatment. Finally, although the most accurate comparison for the normal bow of the radius would be personalized, based on comparison to an individual’s “normal” forearm, the retrospective nature of our data prohibited this evaluation.