Outcomes at a Mean of 13 Years After Proximal Humeral Fracture During Adolescence

Proximal humeral fractures are common sports injuries in adolescents1. The fractures heal well with nonoperative treatment in patients <10 years of age; a surgical procedure is advocated with increasing age and fracture displacement2. Nonoperative treatment is supported by evidence for patients ≤10 years of age; the tremendous remodeling potential of the proximal humerus corrects most osseous deformities, resulting in good to excellent functional outcomes for most children3. However, outcomes in adolescents with lower remodeling potential are poorly understood, as most studies are confounded by the wide age range of the patients2,4–7. Follow-up time has also varied considerably, and, to our knowledge, no studies have performed follow-up of patients to adulthood.

Most proximal humeral fractures in children are treated with immobilization and gradual return to activities8. Surgical treatment options are closed reduction with or without fixation. If adequate reduction cannot be achieved, open reduction with fixation is preferred. Recent trends have favored a surgical procedure, despite the absence of evidence9,10.

We aimed to investigate the course of primary treatment for proximal humeral fractures sustained in adolescence and the long-term upper-limb function and quality of life in adulthood. We compared the outcomes with data from age-matched controls and compared outcomes after operative and nonoperative treatment via propensity score matching.

Materials and Methods Setting, Eligibility, and Participants

This retrospective cohort study was undertaken in the Helsinki University Children’s and Adolescents’ Hospital, a pediatric referral center with a catchment area of approximately 1.5 million people. Approval was obtained from the institutional review board. The study was carried out in accordance with the World Medical Association Declaration of Helsinki. All participants provided written informed consent to participate.

The inclusion criteria were fracture of the humerus proximal to the distal edge of the pectoralis major insertion treated between January 1, 1995, and December 31, 2005; an age of 10.0 to 16.0 years at the time of the injury; and an open physis of the proximal humerus. The exclusion criteria were other concomitant or previous fracture(s) or disease of the shoulder region, pathologic fractures secondary to a malignancy, and conditions affecting the ability to answer study questionnaires.

We searched the hospital database for patients who were treated for a proximal humeral fracture by using the appropriate International Classification of Diseases, Ninth Revision (ICD-9) and Tenth Revision (ICD-10) codes (812.0, 812.1, and S42.2). We reviewed the hospital records and radiographs to identify eligible patients and extracted the demographic and injury details, course and method of treatment, interventions, and harms. Operative management was any intervention requiring the use of operating room resources and included all types of procedures performed in the operating room: closed reduction with or without fracture fixation and open reduction with fixation.

Patients were approached through letters and telephone calls and were asked to attend a follow-up appointment. Those unwilling or unable to visit were invited to participate in a telephone interview. During the follow-up visit, a comprehensive evaluation of the upper extremity was performed by the study physiotherapist, who was aware of which side had been injured in the past, and participants answered the outcome questionnaires. Participants were asked to have a radiograph of the shoulder made by a local provider. Participants with any symptoms were offered an appointment with a shoulder surgeon.

There was no patient or public involvement in this study beyond participation in the follow-up visits.

Outcomes

The primary study outcome was the Disabilities of the Arm, Shoulder and Hand (DASH) questionnaire11. The secondary outcomes were the Simple Shoulder Test (SST)12; pain at rest (using a numeric rating scale [NRS] of 0 to 10, in which higher scores indicate worse pain) and pain with strenuous arm activity (NRS); subjective satisfaction (NRS of 0 to 10, in which higher scores indicate greater satisfaction); individual relative Constant score (IRCS), in which the Constant score of the injured side was divided by the Constant score of the uninjured side and multiplied by 10013; active and passive glenohumeral joint range of motion (flexion, extension, external rotation, and internal and external rotation with the shoulder in 90° of abduction); isometric shoulder abduction strength (as in the strength component of the Constant score); belly-press strength (as in the belly-press test)14, shoulder internal and external rotation strength; olecranon-acromion distance (to represent the length of the humerus); 15D quality-of-life measure15,16; and harms of treatment. Data from the contralateral side were also collected when applicable. Participants who attended the follow-up visit answered the DASH questions. Participants who were interviewed via telephone answered the SST, pain, and subjective satisfaction questions.

The single study radiologist, who was blinded to the outcome data at any point, analyzed the original and follow-up radiographs (reviewed in a single setting) for fracture displacement and angular deformity. The follow-up images were analyzed after the original injury radiographs. The Peterson classification was used to characterize the physeal fractures, with an addition of type 0 for metaphyseal fractures17. The Neer-Horwitz classification was used for fracture displacement18. The long-term radiographs consisted of 2 perpendicular views of the previously fractured humerus to enable the calculation of the maximum true angular deformity.

Statistical Analysis

Logistic regression was used to determine whether any of the potential predictive factors other than fracture characteristics were associated with operative treatment. Participants were compared with age-matched, uninjured controls consisting of patients from the Finnish Health 2011 study (96 patients with a 15D score of 0.9440 [95% confidence interval (CI), 0.9397 to 0.9482; range, 0.93 to 0.99]) and young adults of similar age (mean and standard deviation, 28.8 ± 7.4 years; range, 17 to 50 years) in the United States (192 patients with normal mean shoulder scores of 1.85 [95% CI, 1.0 to 2.7; range, 0 to 50] for the DASH and 11.79 [95% CI, 11.7 to 11.9; range, 9 to 12] for the SST)19,20.

To explore whether our grouping of all procedures requiring anesthesia as operative treatment was justified, we compared data from participants who underwent fracture fixation (after closed or open reduction) with data from participants who underwent closed reduction without fixation. We counted the number of patients who had a change of fracture alignment and compared the outcomes.

Data for harms are presented as observed events with bootstrapped 95% CIs.

The effect of operative treatment was gauged by using propensity score matching to estimate the treatment effect between operatively treated patients and similar nonoperatively treated patients21. Linear regression analyses were performed to identify predictors of inferior results.

Logistic regression was used to compare categorical outcome variables between groups, and linear regression and unequal-variance t tests were used to compare continuous outcome variables. When appropriate, missing values were imputed using the k-nearest neighbor approach22.

Our prediction model for treatment decisions in the full data set was based on information available at the time of the injury: treatment choice, sex, age at the time of the injury, injury energy (high or low), whether the dominant arm was fractured (yes or no), maximum angle measurable in the radiographs at presentation, primary shortening, displacement grade (according to the Neer-Horwitz grade), and fracture type (Peterson class). The full model had 5 potential confounders: follow-up time, subsequent injuries, immobilization method (any hard cast [u-cast, long forearm cast, sugar tong cast] or soft cast [collar and cuff, arm sling, sling and swathe, nothing]), number of radiographs and associated x-ray exposures during the evaluation and follow-up (excluding intraoperative fluoroscopy), and harms. Analysis of variance was used to compare different linear regression models, and the likelihood ratio test was used to compare logistic regression models to choose the most parsimonious model possible.

Propensity score matching used the Matching package in R (R Foundation for Statistical Computing) with the following biologically, clinically, and epidemiologically most plausible variables influencing the choice of treatment modality: age at the time of the injury, sex, follow-up time, primary maximum angulation (degrees), primary maximum displacement (percentage of shaft width), and primary shortening (in centimeters)23.

We anticipated that nonoperative treatment would have a proportionately larger number of patients with very low propensity score (reflecting a nondisplaced fracture). We excluded data from participants with a low score (<0.05) to improve balance and then performed a second propensity matching with the reduced data set. The match balance was evaluated with the omnibus test introduced by Hansen and Bowers24. We matched operatively and nonoperatively treated participants (using the average treatment effect for treated [ATT] and average treatment effect for controls [ATC]) and then calculated the average treatment effect (ATE), the difference between operatively and nonoperatively treated patients25.

We set the level of significance to p < 0.05. Microsoft Excel and R version 3.4.1 software were used for the statistical analyses26.

Source of Funding

There was no external funding for this study.

Results

We included 209 participants (210 fractures). Outcome data were available for 152 participants (153 fractures); 78 participants (78 fractures) attended the follow-up visit, and 62 participants had a radiograph (Fig. 1). The mean follow-up time was 13.1 ± 3.2 years (range, 7.4 to 19.1 years).

fig1Fig. 1:

Study flowchart.

Demographic characteristics and injury mechanisms are presented in Table I, fracture characteristics are presented in Table II, and treatment characteristics are presented in Table III. Of the fractures, 134 (64%) resulted from sport injuries. The most common mechanism of injury was a fall from horseback (n = 41) for girls and a fall during alpine skiing or snowboarding (n = 37) for boys. Seventy-five percent of patients who underwent operative treatment were boys. In the logistic regression model (176 patients with sufficient data), sex was not an independent predictor of treatment modality (p = 0.14). Choosing operative treatment was predicted by increasing primary fracture displacement (p < 0.001) and angulation (p = 0.009) and by a Peterson type-II fracture (p = 0.016). Data according to follow-up status and treatment group are shown in Appendix Table 1.

TABLE I - Demographic Characteristics of the Fractures* No. of fractures 210 Age at the time of injury†(yr) 12.9 ± 1.6 (10.0 to 16.0) Age at follow-up†(yr) 26.0 ± 3.4 (18.5 to 33.6) Female sex‡ 106 (50%) Dominant arm injured‡§ 31 (40%) Nonsurgical treatment‡ 164 (78%) Mechanisms of injury  Alpine skiing or snowboarding# 49   Girls 12   Boys 37  Fall from horseback# 42   Girls 41   Boys 1  Simple fall 27   Girls 18   Boys 9  Fall from height (<3 m) (usually a tree) 20   Girls 10   Boys 10  Ice hockey or skating# 13   Girls 5   Boys 8  Pedestrian in motor vehicle accident 12   Girls 5   Boys 7  Biking and moped accidents 10   Girls 3   Boys 7  Skateboarding# 8   Girls 1   Boys 7  Other sports injuries# (e.g., gymnastics, football) 22   Girls 10   Boys 12  Other** 7   Girls 1   Boys 6 Sports injuries‡ 134 (64%)

*One girl fractured both shoulders at different times.

†The values are given as the mean and the standard deviation, with the range in parentheses.

‡The values are given as the number of fractures, with the percentage in parentheses.

§Arm dominance was reliably available from the 78 participants who attended the follow-up visit.

#These injury mechanisms were classified as sports injuries.

**This category includes motor vehicle accidents, falls in special circumstances, and violence.


TABLE II - Fracture Characteristics in Patients with Available Radiographs* Characteristic Nonoperative Treatment (N = 145†) Operative Treatment (N = 37†) No. of Fractures Angulation‡(deg) Displacement‡(% of bone width) No. of Fractures Angulation‡(deg) Displacement‡(% of bone width) Fracture type  Metaphyseal 87 17 ± 12 (0 to 60) 11 ± 23 (0 to 100) 12 33 ± 13 (14 to 53) 90 ± 33 (29 to 100)  Peterson type   I 7 15 ± 13 (0 to 35) 10 ± 8 (0 to 23) 0 — —   II 47 22 ± 11 (3 to 52) 24 ± 21 (0 to 100) 20 33 ± 12 (15 to 66) 55 ± 32 (0 to 100)   III 3 32 ± 11 (23 to 47) 31 ± 13 (13 to 43) 4 51 ± 24 (15 to 83) 75 ± 26 (41 to 100)  Other 1§ 7 100 1# 33 70 Neer-Horwitz grade   III** 14†† 11‡‡   IV§§ 9†† 20‡‡

*Values are given for fractures.

†Number of radiographs available.

‡The values are given as the mean and the standard deviation, with the range in parentheses. §This was a fracture through a simple bone cyst.

#This was a complex multifragmentary fracture.

**Displacement of more than one-third up to equal to two-thirds of the bone width.

††Data were available from 143 fractures.

‡‡Data were available from 35 fractures.

§§Displacement of more than two-thirds of the bone width.


TABLE III - Fracture Treatment Characteristics Assessed from Patients’ Medical Records Characteristic Nonoperative Treatment (N = 146) Operative Treatment (N = 46) Immobilization*  Collar and cuff 93 20  Hanging cast 31 7  Arm sling 22 13  Other cast 7 6  Other 8 0 Duration of immobilization†(wk) 3.0 ± 0.2 (1 to 3) 4.0 ± 1.0 (1 to 6) Weeks until free use of arm after the injury† 5.6 ± 1.9 (0 to 14) 7.4 ± 1.9 (4 to 12) Fixation* — 34

*The values were given as the number of fractures; some fractures had >1 type of immobilization.

†The values were given as the mean and the standard deviation, with the range in parentheses.


Primary Outcome

The mean DASH score was 2.5 (95% CI, 1.8 to 3.3 [range, 0 to 13]) among the 74 participants with data. The distribution of DASH scores is presented in Figure 2-A. There were no clinically important or significant differences (p = 0.38) in the DASH scores compared with the normal values in healthy adults (1.85 [95% CI, 1.0 to 2.7]) (Table IV). Outcomes were very similar in participants treated operatively and nonoperatively (see Appendix Table 2).

fig2Fig. 2:

Figs. 2-A through 2-F Distribution of outcome scores. Fig 2-A DASH scores, in which lower scores indicate better function. Fig 2-B Pain scores at rest, in which 0 indicates no pain and 10 indicates the worst imaginable pain. Fig. 2-C IRCS, in which the injured side is divided by the uninjured side and multiplied by 100; 100 indicates the same Constant score on both sides. Fig. 2-D Pain scores with activity, in which 0 indicates no pain and 10 indicates the worst imaginable pain. Fig. 2-E Satisfaction scores, in which 0 indicates very dissatisfied and 10 indicates fully satisfied. Fig. 2-F SST scores, in which higher scores indicate better function.

TABLE IV - Patient-Reported and Other Outcomes Outcome Study Patients Normal Population Reference Values* P Value No. of Fractures Mean* Primary outcome  DASH† 74 3 (2 to 3) 2 (1 to 3) 0.38 Secondary outcomes  SST‡ 153 11.7 (11.5 to 11.8) 11.8 (11.7 to 11.9) 0.23  Pain at rest§ 149 0.1 (0 to 0.3)  Pain in strenuous activity§ 150 0.7 (0.5 to 0.9)  Subjective satisfaction# 150 9 (8.8 to 9.3)  IRCS**(%) 78 98 (96 to 99)  15D†† 75 0.964 (0.953 to 0.975) 0.944 (0.940 to 0.948) <0.001

*The values are given as the mean, with the 95% CI in parentheses.

†Scored from 0 to 100; a higher score is a worse result.

‡Scored from 0 to 12; a higher score is a better result.

§On an NRS of 0 to 10; a higher score is a worse result.

#On an NRS of 0 to 10; a higher score is a better result.

**Individual relative Constant score; 100% indicates equal Constant scores on both sides.

††Scored from 0 to 1; a higher score is a better result.


Secondary Outcomes

The secondary outcomes are presented in Table V. There were no significant differences in the SST scores compared with the normal values in healthy adults. There was a significant difference in the glenohumeral joint range of motion compared with the uninjured side: 5° in active flexion and active abduction and 8° in active external rotation in 90° of abduction. There were no differences in other outcomes compared with the uninjured side. Distributions of pain scores, satisfaction, SST scores, IRCS, and subjective satisfaction are shown in Figures 2-B through 2-F.

TABLE V - Comparison of the Injured and Uninjured Sides (N = 78) Outcome Injured Side* Uninjured Side* P Value Active flexion range of motion (deg) 159 (156 to 162) 164 (162 to 166) 0.02 Active abduction range of motion (deg) 170 (167 to 174) 176 (175 to 178) 0.01 Active external rotation with arm at side (deg) 78 (75 to 80) 81 (78 to 83) 0.13 Active external rotation at 90° abduction (deg) 83 (80 to 86) 91 (88 to 93) <0.001 Active internal rotation at 90° abduction (deg) 81 (78 to 84) 83 (80 to 85) 0.38 Abduction strength, as in Constant score (kg) 5.6 (5.1 to 6.1) 5.6 (5.2 to 6.1) 1.00 Belly-press strength (kg) 6.6 (6.1 to 7.1) 6.4 (6.0 to 6.9) 0.67 External rotation strength (kg) 5.6 (5.2 to 6.1) 6.1 (5.1 to 7.1) 0.37 Internal rotation strength (kg) 7.2 (6.6 to 7.8) 7.2 (6.7.7 to 7.8) 0.94 Olecranon-acromion distance (cm) 35.1 (34.6 to 35.7) 35.4 (34.9 to 36.0) 0.41

*The values are given as the mean, with the 95% CI in parentheses.

The mean 15D score was 0.964 (95% CI, 0.953 to 0.975). Compared with the reference population, the participants had a higher quality of life as measured by the 15D instrument; the difference was significant and clinically important15,16 (mean difference, 0.020 [95% CI, 0.006 to 0.031]; p < 0.001) (Fig. 3).

fig3Fig. 3:

Quality-of-life 15D instrument domains and total score in the age-matched general population and participants in our study.

Of the 62 participants with radiographs, 36 had normal results with no angular deformity. Among the remaining 26 participants, the mean maximum angular deformity was 13.6° (95% CI, 10.9° to 16.3°). Two participants had >30° of maximum angular deformity; 1 had a DASH score of 0 and an SST score of 12, and the other had a DASH score of 13 and an SST score of 8.

In the regression modeling, we could not identify a predictor of inferior DASH, SST, pain, satisfaction, or IRCS outcomes. The adjusted R2 values were low (<0.15).

The results of outcomes at long-term follow-up according to treatment choice and follow-up status are presented in Appendix Table 2.

Operative Treatment with or without Fixation

Long-term follow-up data were available for 9 participants who underwent manipulation under anesthesia without fixation and 23 participants who underwent fracture fixation (all with percutaneous pinning). There were no significant differences in outcomes between these subgroups. Among the 11 patients who did not undergo fixation but had follow-up radiographs from the primary treatment episode, 8 maintained alignment in the fractures during follow-up, 2 improved (both had Peterson type 0), and 1 had angulation return to the pre-manipulation degree (Peterson type III). All 27 patients who underwent fracture fixation and had available follow-up radiographs maintained the fracture alignment achieved at the time of the surgical procedure.

Harms

Of the operatively treated patients, 4 (12% [95% CI, 2% to 24%]) had a superficial pin-track infection and 1 (3% [95% CI, 0% to 9%]) had a deep infection. These did not require a revision surgical procedure. One participant (3% [95% CI, 0% to 9%]) reported painless shoulder stiffness (with measured flexion of 110°) at the long-term follow-up.

Of the nonoperatively treated patients, 1 (1% [95% CI, 0% to 2%]) had shoulder stiffness, which did not resolve. The follow-up radiograph was normal.

Propensity Matching

We used data from the 78 participants who attended the in-person follow-up for propensity score matching for the primary outcome. Before matching, the operatively and nonoperatively treated participants were not comparable (p = 0.004). Data from 15 nonoperatively treated participants were excluded from the analysis due to a low propensity score, reflecting a nondisplaced fracture. After 38 of the remaining 63 patients were successfully matched, the omnibus test p values for balance were 0.805 for ATT, 0.0002 for ATC, and 0.0008 for ATE, indicating that we were able to balance the data for the ATT analysis but not for the ATE and ATC analyses. There were no significant differences in any outcomes between operative and nonoperative treatment (Table VI). As a sensitivity analysis, we used data from the 152 respondents (with 153 fractures) with any follow-up data and repeated the propensity score matching analysis, with a similar result of no difference in outcomes between operative and nonoperative treatment (Table VII).

TABLE VI - ATT in the 19 Propensity-Matched Pairs Among the 78 Patients with Full Follow-up Outcome No. of Patients Between-Group Difference* P Value for Null Hypothesis† Primary outcome  DASH‡ 38 1 (−1 to 3) 0.24 Secondary outcomes  SST§ 38 −0.2 (−1.0 to 0.6) 0.60  Pain at rest# 38 0.2 (−0.2 to 0.6) 0.27  Pain in strenuous activity# 38 −0.3 (−1.7 to 1.1) 0.70  Subjective satisfaction** 38 −0.3 (−1.2 to 0.7) 0.59  IRCS††(%) 38 −3 (−7 to 0) 0.07

*The ATT values (average treatment effect for the treated group, calculated as the operative group minus the nonoperative group) are given as the mean difference, with the 95% CI in parentheses.

†The probability of no true difference between groups after matching. There was no significant difference between groups after matching.

‡Scored from 0 to 100; a higher score is a worse result.

§Scored from 0 to 12; a higher score is a better result.

#On an NRS of 0 to 10; a higher score is a worse result.

**On an NRS of 0 to 10; a higher score is a better result.

††Individual relative Constant score; 100% indicates equal Constant scores on both sides.


TABLE VII - ATT in the 32 Propensity-Matched Pairs of Fractures Among the 152 Patients with Any Follow-up Outcome No. of Patients Between-Group Difference* P Value for Null Hypothesis†

留言 (0)

沒有登入
gif