Fibular allograft augmentation shows potential benefits in treating medial column comminuted proximal humeral fractures with osteoporosis: a secondary analysis of a randomized controlled trial Authors: Yijie Chai, Tingwang Shi, Haochen Jiang, Yujie Chen, Zhengliang Xu, Yu Zhu, Zesong Ruan, Yunfeng Chen, Qiuke Wang Published: 2026 ABSTRACT BackgroundFibular allograft (FA) augmentation is proposed to address the lack of medial column support in plate-fixed proximal humeral fractures; however, the role of FA augmentation in these fractures remains inadequately elucidated. Our previous clinical trial did not show additional benefit in inserting an FA when treating adults with medial column comminuted proximal humeral fractures. Nonetheless, it is possible that some subgroups may still derive benefits from FA. This study aimed to investigate whether there were any subgroups benefit from FA.Materials and methodsOur randomized controlled trial (RCT) recruited adults with a medial column comminuted proximal humeral fracture and randomly allocated them to receive a surgery fixed with a locking plate (LP group) or with an LP augmented with an FA (FA group). This study was a subsequent subgroup analysis of the previous randomized controlled trial, in which we included 72 participants who completed follow-up at the time of 12 months for the current subgroup analysis. The primary outcome was the Disabilities of the Arm, Shoulder and Hand (DASH) score at 12 months postoperatively. The secondary outcomes included the DASH score at other time points, changes in neck-shaft angle and humeral head height. We used generalized linear models for subgroup analysis to evaluate the effect of intervention (FA or LP) on 12-month DASH score in subgroups.ResultsA total of 72 participants who completed follow-up at 12 months were included: 35 were in the FA group and 37 in the LP group. In participants with proximal humeral osteoporosis, both the unadjusted mean difference (−12.8; 95% confidence interval (CI) −22.2 to −3.2; p = 0.01) and the adjusted mean difference (−11.1; 95% CI −20.0 to −2.2; p = 0.02) in the 12-month DASH score were statistically significant. Furthermore, in proximal humeral osteoporosis category, the DASH scores at other time points showed the similar trends though did not reach statistical significance.ConclusionsFA augmentation seemed to have added values in treating osteoporotic proximal humeral fractures in combination with medial column comminution.Level of evidenceLevel II, therapeutic study.Trial registrationChiCTR-IOR-16008817. Introduction Proximal humeral fractures have been estimated to account for 5–6% of all adult fractures, making them the third most common osteoporotic fracture type [1, 2]. Approximately 25% of proximal humeral fractures require surgical treatment [3]. Open reduction and internal fixation (ORIF) with a locking plate (LP), which helps achieve osteosynthesis, is frequently indicated for displaced proximal humeral fractures [4]. Although locking plate fixation can provide promising clinical outcomes, it may also lead to challenging complications such as secondary intra-articular screw penetration, varus collapse, and humeral head necrosis [5, 6]. Lack of medial column support has been described by some studies as an important factor in the development of these complications [7, 8]. To overcome these, Gardner et al. suggested using a fibular allograft (FA) strut, along with a laterally placed locking plate, as structural augmentation to provide additional medial column support [9, 10]. The use of fibular allograft offers both advantages and disadvantages in terms of its application. Primarily, the fibular allograft provides essential structural support and mechanical stability to the medial column, which is crucial for effective fracture fixation. This technique can significantly reduce the risk of fixation failure and the need for revision procedures, particularly in complex or challenging fractures, but it also introduces potential drawbacks, such as risks of infection, higher costs, limited availability, increased surgical complexity, and the possibility of approach-related complications [9, 10]. A recent randomized controlled trial, performed by our team, did not find additional benefits found in augmentation with an FA when treating medial column comminuted proximal humeral fractures [11]. This is different from the results of several studies [12–16]. For example, a meta-analysis indicated that participants with proximal humeral fractures treated with FA augmentation would achieve improved clinical outcomes; a comparative study showed that in participants with osteoporosis, the FA group had better forward elevation function and fewer postoperative complications [12, 16]. It should be noted that these findings were based on retrospective studies, which introduces methodological biases. Another possible reason could be that there were subgroups who benefited from FA, but the effects in these subgroups were diluted in the primary analysis of the previous trial. Thus, this study was a subsequent subgroup analysis of our previous randomized clinical trial, aiming to investigate whether there were any subgroups benefit from FA. Materials and methods Participants This study is a post hoc subgroup analysis of a single-center, open-label with respect to surgeons and participants (function assessor and data analyst-blinded), two-arm (1:1), parallel-group, superiority randomized controlled trial [11]. The recruitment period for the RCT spanned from 20 October 2016 to 24 December 2019. The inclusion criteria were as follows: aged 18 years and over; unilateral proximal humeral fracture involving medial column comminution (medial column comminution was defined as a fracture of the medial column with at least three fragments, according to established criteria in previous studies) [11]; acute fracture (within 3 weeks); ORIF surgery is indicated; and the patient agrees and gives written informed consent. Exclusion criteria included fractures in other regions, fractures extending into the humeral shaft, a history of previous surgical treatment for ipsilateral upper-extremity fractures, chronic shoulder arthritis, open fractures, and pathological fractures, such as those related to tumors, osteomalacia, osteomyelitis, or simple bone cysts. For the current study, we included participants who completed follow-up at the time of 12 months from the previous RCT study, because the previous RCT used the DASH scores at 12 months as the primary outcome and 12 months provided sufficient time to evaluate the stability and durability of the treatment effect. Interventions Participants were randomly allocated to undergo ORIF with an LP augmented with an FA (the FA group) or an LP alone (the LP group). Two surgeons, each with over 20 years of experience in shoulder surgery, performed all procedures. In the FA group, participants were positioned in the beach chair posture and administered general anesthesia. Surgeons used a deltopectoral approach to access and reduce the fracture fragments with the aid of Kirschner wires and a fibular allograft for medial support. The FA was inserted into the distal medullary canal and directed toward the medial calcar to indirectly reduce the medial column and provide support to the humeral head fragment. Once satisfactory fracture reduction was achieved, a locking plate (LP) (Philos; DePuy Synthes) was applied lateral to the bicipital groove on the humerus. If additional support was required, cancellous bone grafts from the ipsilateral iliac crest or bone substitutes were utilized, with each instance documented accordingly. After securing the fracture, the wound was thoroughly irrigated and closed in layers. In the LP group, the surgical procedures were nearly identical to those in the FA group, with the key difference being that the fractures were treated solely with the LP (Philos) without the use of an intramedullary FA. Subgroups We created subgroups based on baseline factors: sex, age, body mass index (BMI), current alcohol drinker, current smoker, hypertension, diabetes, Neer classification, AO/OTA classification, participant-reported osteoporosis, and proximal humeral osteoporosis. In our previous study, participant-reported osteoporosis was defined as the disease history of participants who said they had been diagnosed as osteoporosis in the outpatient clinic. Due to the fact that whole-body bone density testing does not accurately represent the bone density of the proximal humerus and also to the possibility of recall bias, we also used the Tingart index to define proximal humeral osteoporosis in this study [17]. The Tingart index measured the cortical thickness of the proximal humeral diaphysis at two distinct levels (Fig. 2). Participants with a Tingart index less than 4 mm were considered to have proximal humeral osteoporosis. To ensure measurement reliability and reduce bias, all Tingart index assessments were performed independently by two reviewers who were blinded to treatment allocation and clinical outcomes. We divided age category into two stratifications: 18 to 64 years and ≥ 65 years (as indicated by previous literature), and BMI into two stratifications: 19.0 to 25.0 and ≥ 25.0(no participants with BMI < 19.0 in this study) [17]. Outcomes We used the same primary outcome for this study as the original trial that is Disabilities of the Arm, Shoulder, and Hand (DASH) questionnaire score measured at 12 months after the surgery [18]. The DASH questionnaire scored from 0 to 100, where 0 indicates the absence of symptoms or disabilities. To verify the findings from the primary results, we used the secondary outcomes which included the DASH score at 3 and 6 months and radiographic outcomes. The radiographic outcomes were the changes in neck-shaft angle (NSA) and the humeral head height (HHH) from the first day after the surgical procedure. Statistical analysis We described the baseline information of participants, using mean (standard deviation) for continuous variables and the number of participants (percentage) for categorical variables. To study intergroup heterogeneity, we first used generalized linear models to assess the effect of the interaction between subgroup variables and the intervention (FA or LP) on 12-month DASH scores. The model structure included the intervention as the primary exposure, subgroup as a categorical variable, and an intervention × subgroup interaction term to test for heterogeneity of treatment effect across subgroups. We then used generalized linear models for subgroup analysis to evaluate the effect of intervention (FA or LP) on 12-month DASH score in subgroups. In these models, the intervention (FA versus LP) was the independent variable, and the 12-month DASH score was the dependent variable. To mitigate the impact of confounding bias, we have adjusted our analysis to account for potential confounding variables. Covariates were selected based on clinical relevance and observed baseline imbalances. For example, we adjusted for BMI, Neer classification, and quality of reduction in the subgroup of proximal humeral osteoporosis. We calculated the interaction p value, the unadjusted and adjusted mean differences, and their 95% confidence intervals (CIs) between the two interventions within subgroups. Significance was set at p < 0.05. According to Gummesson’s research, a minimal clinically important difference (MCID) in mean DASH score was ten-point difference [19]. We selected the subgroups in which the mean difference between the two intervention measures was greater than or equal to MCID (and FA could bring better clinical outcomes in this subgroup) and p < 0.05 at the same time for further result validation. Next, we used generalized linear models for result validation to evaluate the effect of intervention (FA or LP) on 3- and 6-month DASH scores in chosen subgroups. We divided the participants into four groups according to selected subgroup and intervention. Then we analyzed the change trend of DASH score, the changes in neck shaft angle (NSA), and the changes in head height of humerus (HHH) over time. Significance was set at p < 0.05. For the missing values of participants’ DASH scores and radiographic outcomes at other time points (selected participants all had complete data at 12 months), we applied multiple imputation by creating five imputed datasets for the subsequent results validation. To test the robustness of our results, we performed sensitivity analysis on the complete cases at each follow-up time point. Results Participants Eight participants were excluded due to the lack of a 12-month follow-up. A total of 72 participants who were followed up to 12 months were included. Among them, 35 were in the FA group and 37 were in the LP group. See the detailed process in Fig. 1. The mean age (standard deviation), number of females, and mean BMI (standard deviation) of FA group were 64 (12) years, 21 (60%), and 25.4 (3.5), respectively; 66 (15) years, 26 (70%), and 24.8 (3.2) of LP group. The detailed baseline characteristics are presented in Table 1 (Fig. 2).Fig. 1Flowchart of patient recruitment and follow-upTable 1Baseline and surgical characteristics of the population who completed 12 months of follow-upCharacteristicFA group (N = 35)LP group (N = 37)Age*(years), mean (SD)64 (12)66 (15)Female sex†, n (%)21 (60)26 (70)BMI*(kg/m2), mean (SD)25.4 (3.5)24.8 (3.2)Fracture on the left side†, n (%)16 (46)17 (46)Left arm dominant†, n (%)6 (17)10 (27)Cause of the fracture†–– Fall on the ground‡, n (%)29 (83)33 (89) Fall on the stairs, n (%)0 (0)1 (3) Traffic accident, n (%)6 (17)3 (8)Current smoker†, n (%)11 (31)4 (11)Current alcohol drinker†, n (%)13 (37)9 (24)Comobidities†–– Diabetes, n (%)8 (23)4 (11) Hypertension, n (%)7 (20)8 (22) Previous fracture, n (%)2 (6)4 (11) Participant-reported osteoporosis, n (%)12 (34)13 (35)Neer classification† §–– Two part, n (%)9 (26)9 (24) Three part, n (%)12 (34)16 (43) Four part, n (%)14 (40)12 (32)AO/OTA classification† # **–– Type A, n (%)9 (26)8 (22) Type B, n (%)12 (34)14 (38) Type C, n (%)14 (40)15 (41)Fracture with shoulder dislocation†, n (%)4 (11)1 (3)VAS pain score*, mean (SD)6.6 (1.9)6.3 (1.7)Time from fracture to surgery (days)*, mean (SD)4.1 (2.6)3.5 (1.7)Use of cancellous bone graft† ††, n (%)11 (31)12 (32)Good reduction†, n (%)19 (54)24 (65)*The values are given as the mean and the standard deviation†The values are given as the number of participants, with the percentage in parentheses‡Includes falls while standing, walking, jogging, and doing other sports§Graded by two independent reviewers with an intraclass correlation coefficient (ICC) of 0.94 (95% CI 0.91 to 0.96)#Graded by two independent reviewers with an ICC of 0.94 (95% CI 0.91 to 0.96)**The OTA/AO classification system††Cancellous allogeneic bone substitutes; no autologous bone grafting (iliac crest) was performedFig. 2The lateral and medial cortices were measured at levels 1 and 2. Level 1 represented the most proximal segment of the humeral diaphysis where the endosteal boundaries of the lateral and medial cortices ran parallel to each other. Level 2 was situated 20 mm distal to level 1. The Tingart index was defined as the average of the medial and lateral cortical thicknesses at these two levels Flowchart of patient recruitment and follow-up Table 1Baseline and surgical characteristics of the population who completed 12 months of follow-upCharacteristicFA group (N = 35)LP group (N = 37)Age*(years), mean (SD)64 (12)66 (15)Female sex†, n (%)21 (60)26 (70)BMI*(kg/m2), mean (SD)25.4 (3.5)24.8 (3.2)Fracture on the left side†, n (%)16 (46)17 (46)Left arm dominant†, n (%)6 (17)10 (27)Cause of the fracture†–– Fall on the ground‡, n (%)29 (83)33 (89) Fall on the stairs, n (%)0 (0)1 (3) Traffic accident, n (%)6 (17)3 (8)Current smoker†, n (%)11 (31)4 (11)Current alcohol drinker†, n (%)13 (37)9 (24)Comobidities†–– Diabetes, n (%)8 (23)4 (11) Hypertension, n (%)7 (20)8 (22) Previous fracture, n (%)2 (6)4 (11) Participant-reported osteoporosis, n (%)12 (34)13 (35)Neer classification† §–– Two part, n (%)9 (26)9 (24) Three part, n (%)12 (34)16 (43) Four part, n (%)14 (40)12 (32)AO/OTA classification† # **–– Type A, n (%)9 (26)8 (22) Type B, n (%)12 (34)14 (38) Type C, n (%)14 (40)15 (41)Fracture with shoulder dislocation†, n (%)4 (11)1 (3)VAS pain score*, mean (SD)6.6 (1.9)6.3 (1.7)Time from fracture to surgery (days)*, mean (SD)4.1 (2.6)3.5 (1.7)Use of cancellous bone graft† ††, n (%)11 (31)12 (32)Good reduction†, n (%)19 (54)24 (65)*The values are given as the mean and the standard deviation†The values are given as the number of participants, with the percentage in parentheses‡Includes falls while standing, walking, jogging, and doing other sports§Graded by two independent reviewers with an intraclass correlation coefficient (ICC) of 0.94 (95% CI 0.91 to 0.96)#Graded by two independent reviewers with an ICC of 0.94 (95% CI 0.91 to 0.96)**The OTA/AO classification system††Cancellous allogeneic bone substitutes; no autologous bone grafting (iliac crest) was performed Baseline and surgical characteristics of the population who completed 12 months of follow-up *The values are given as the mean and the standard deviation †The values are given as the number of participants, with the percentage in parentheses ‡Includes falls while standing, walking, jogging, and doing other sports §Graded by two independent reviewers with an intraclass correlation coefficient (ICC) of 0.94 (95% CI 0.91 to 0.96) #Graded by two independent reviewers with an ICC of 0.94 (95% CI 0.91 to 0.96) **The OTA/AO classification system ††Cancellous allogeneic bone substitutes; no autologous bone grafting (iliac crest) was performed The lateral and medial cortices were measured at levels 1 and 2. Level 1 represented the most proximal segment of the humeral diaphysis where the endosteal boundaries of the lateral and medial cortices ran parallel to each other. Level 2 was situated 20 mm distal to level 1. The Tingart index was defined as the average of the medial and lateral cortical thicknesses at these two levels Subgroup analysis The results of the subgroup analyses are summarized in Table 2. The DASH_12m Difference was obtained by subtracting the LP group DASH Score from the FA group DASH Score. For the proximal humeral osteoporosis subgroups, FA augmentation obtained better results than using LP alone, with statistically significant unadjusted mean difference (−12.8 [95% CI −22.2 to −3.2]; p = 0.01), adjusted mean difference (−11.1 [95% CI −20.0 to −2.2]; p = 0.02) and a significant interaction (p for interaction < 0.001). For the patient-reported osteoporosis subgroups, the inter-group differences beyond the MCID and showed similar trend toward favoring FA augmentation, though the unadjusted and adjusted mean differences (−11.8 [95% CI −24.1 to 0.62]; −16.2 [95% CI −34.3 to 1.9]) was not statistically significant. For the alcohol drinking subgroups, there was a statistically significant difference which seemed to show a similar trend toward favoring FA augmentation in the unadjusted mean difference (13.6 [95% CI −24.5 to −2.7]; p = 0.04), with a significant interaction (p for interaction = 0.01). Table 2Results of the generalized linear models for subgroup analysis in 12-month DASH scoreVariableSubgroup*FA group DASH score&LP group DASH score&Dash score difference at 12 months†p valueAdjusted dash score difference at 12 months†§p value§Interaction p valueSexMale (n = 25)17.3(3.4)16.7(3.9)0.5(−9.6 to 10.6)0.925.9(−2.6 to 14.5)0.180.74Female (n = 47)19.1(2.8)20.7(2.5)−1.6(−9.1 to 5.8)0.67−3.0(−10 to 3.9)0.39Age < 65 (n = 26)13.7(3.2)9.6(3.4)4.0(−5.1 to 13.2)0.392.7(−6 to 11.3)0.550.24 ≥ 65 (n = 46)21.5(2.5)24.3(2.3)−2.8(−9.5 to 4.0)0.42−2.0(−8.4 to 4.3)0.53BMI < 25 (n = 39)20.7(3.2)19.4(3.1)1.3(−7.5 to 10.1)0.783.1(−4.3 to 10.4)0.420.38 ≥ 25 (n = 33)15.6(2.8)19.6(2.8)−4.0(−11.8 to 3.7)0.31−4.8(−12.8 to 3.1)0.23DiabetesWithout (n = 60)15.2(2.4)19.8(2.1)−4.6(−10.9 to 1.6)0.14−4.3(−10.1 to 1.4)0.140.04With (n = 12)28.9(4.5)16.7(6.4)12.2(−3.2 to 27.6)0.1212.3(0.7 to 23.9)0.04HypertensionWithout (n = 57)17.3(2.5)18.0(2.4)−0.7(−7.5 to 6.1)0.84−0.4(−6.4 to 5.7)0.910.81With (n = 15)22.5(4.4)24.9(4.1)−2.4(−14.3 to 9.5)0.69−4.6(−14.5 to 5.2)0.36Alcohol drinkingWithout (n = 50)21.0(2.6)16.9(2.3)4.1(−2.7 to 10.8)0.243.1(−3.2 to 9.4)0.330.01**With (n = 22)13.8(3.6)27.5(4.3)−13.6(−24.5 to −2.7)0.01**−7.4(−19.1 to 4.4)0.22SmokingWithout (n = 57)19.4(2.7)19.6(2.3)−0.3(−7.1 to 6.6)0.94−1.2(−7.5 to 5.2)0.720.78With (n = 15)16.2(3.8)18.7(6.3)−2.5(−17.0 to 12.0)0.739.7(−5.6 to 25)0.22Neer classificationTwo part (n = 18)22.2(3.8)10.4(3.8)11.8(1.2 to 22.4)0.03**11.0(4.1 to 17.9)0.00**0.03**Three part (n = 28)14.3(3.5)22.1(3.0)−7.8(−16.9 to 1.2)0.09−8.5(−17.3 to 0.3)0.06Four part (n = 26)19.4(3.4)22.9(3.7)−3.5(−13.4 to 6.4)0.49−2.8(−12.3 to 6.6)0.55OTA/AO classificationType A (n = 17)21.3(3.6)11.2(3.8)10.1(−0.3 to 20.5)0.06**10.8(3.4 to 18.3)0.00**0.10Type B (n = 26)15.0(3.8)19.1(3.5)−4.1(−14.3 to 5.9)0.42−4.0(−13.3 to 5.3)0.40Type C (n = 28)19.4(3.3)24.3(3.2)−5.0(−14.0 to 4.2)0.29−3.8(−12.3 to 4.8)0.39Proximal humeral osteoporosisWithout (n = 47)18.3(2.3)13.2(2.3)5.1(−1.2 to 11.5)0.115.7(−0.3 to 11.7)0.06 < 0.001**With (n = 25)18.4(3.5)31.1(3.4)−12.8(−22.2 to −3.2)0.01**−11.1(−20.1 to −2.2)0.02**Participant-reported osteoporosisWithout (n = 62)17.7(2.5)18.3(2.3)−0.66(−7.2 to 5.9)0.85−0.42(−6.7 to 5.9)0.900.23With (n = 10)21.0(3.5)32.8(5.3)−11.8(−24.1 to 0.62)0.06−16.2(−34.3 to 1.9)0.08*Numbers in parentheses represent the number of participants in this subgroup**indicates statistically significant results (p < 0.05)&These are the observed values given as the mean (SD)†The FA group was the reference. The values are given as the mean, calculated on the basis of marginal means, with the 95% CI in parentheses§For Neer and OTA/AO classifications, we adjusted for baseline BMI, quality of reduction, and proximal humeral osteoporosis. For participant-reported osteoporosis and proximal humeral osteoporosis, we adjusted for baseline BMI, quality of reduction, and Neer classification. For the remaining subgroup variables, we adjusted for proximal humeral osteoporosis, quality of reduction, and Neer classification Table 2Results of the generalized linear models for subgroup analysis in 12-month DASH scoreVariableSubgroup*FA group DASH score&LP group DASH score&Dash score difference at 12 months†p valueAdjusted dash score difference at 12 months†§p value§Interaction p valueSexMale (n = 25)17.3(3.4)16.7(3.9)0.5(−9.6 to 10.6)0.925.9(−2.6 to 14.5)0.180.74Female (n = 47)19.1(2.8)20.7(2.5)−1.6(−9.1 to 5.8)0.67−3.0(−10 to 3.9)0.39Age < 65 (n = 26)13.7(3.2)9.6(3.4)4.0(−5.1 to 13.2)0.392.7(−6 to 11.3)0.550.24 ≥ 65 (n = 46)21.5(2.5)24.3(2.3)−2.8(−9.5 to 4.0)0.42−2.0(−8.4 to 4.3)0.53BMI < 25 (n = 39)20.7(3.2)19.4(3.1)1.3(−7.5 to 10.1)0.783.1(−4.3 to 10.4)0.420.38 ≥ 25 (n = 33)15.6(2.8)19.6(2.8)−4.0(−11.8 to 3.7)0.31−4.8(−12.8 to 3.1)0.23DiabetesWithout (n = 60)15.2(2.4)19.8(2.1)−4.6(−10.9 to 1.6)0.14−4.3(−10.1 to 1.4)0.140.04With (n = 12)28.9(4.5)16.7(6.4)12.2(−3.2 to 27.6)0.1212.3(0.7 to 23.9)0.04HypertensionWithout (n = 57)17.3(2.5)18.0(2.4)−0.7(−7.5 to 6.1)0.84−0.4(−6.4 to 5.7)0.910.81With (n = 15)22.5(4.4)24.9(4.1)−2.4(−14.3 to 9.5)0.69−4.6(−14.5 to 5.2)0.36Alcohol drinkingWithout (n = 50)21.0(2.6)16.9(2.3)4.1(−2.7 to 10.8)0.243.1(−3.2 to 9.4)0.330.01**With (n = 22)13.8(3.6)27.5(4.3)−13.6(−24.5 to −2.7)0.01**−7.4(−19.1 to 4.4)0.22SmokingWithout (n = 57)19.4(2.7)19.6(2.3)−0.3(−7.1 to 6.6)0.94−1.2(−7.5 to 5.2)0.720.78With (n = 15)16.2(3.8)18.7(6.3)−2.5(−17.0 to 12.0)0.739.7(−5.6 to 25)0.22Neer classificationTwo part (n = 18)22.2(3.8)10.4(3.8)11.8(1.2 to 22.4)0.03**11.0(4.1 to 17.9)0.00**0.03**Three part (n = 28)14.3(3.5)22.1(3.0)−7.8(−16.9 to 1.2)0.09−8.5(−17.3 to 0.3)0.06Four part (n = 26)19.4(3.4)22.9(3.7)−3.5(−13.4 to 6.4)0.49−2.8(−12.3 to 6.6)0.55OTA/AO classificationType A (n = 17)21.3(3.6)11.2(3.8)10.1(−0.3 to 20.5)0.06**10.8(3.4 to 18.3)0.00**0.10Type B (n = 26)15.0(3.8)19.1(3.5)−4.1(−14.3 to 5.9)0.42−4.0(−13.3 to 5.3)0.40Type C (n = 28)19.4(3.3)24.3(3.2)−5.0(−14.0 to 4.2)0.29−3.8(−12.3 to 4.8)0.39Proximal humeral osteoporosisWithout (n = 47)18.3(2.3)13.2(2.3)5.1(−1.2 to 11.5)0.115.7(−0.3 to 11.7)0.06 < 0.001**With (n = 25)18.4(3.5)31.1(3.4)−12.8(−22.2 to −3.2)0.01**−11.1(−20.1 to −2.2)0.02**Participant-reported osteoporosisWithout (n = 62)17.7(2.5)18.3(2.3)−0.66(−7.2 to 5.9)0.85−0.42(−6.7 to 5.9)0.900.23With (n = 10)21.0(3.5)32.8(5.3)−11.8(−24.1 to 0.62)0.06−16.2(−34.3 to 1.9)0.08*Numbers in parentheses represent the number of participants in this subgroup**indicates statistically significant results (p < 0.05)&These are the observed values given as the mean (SD)†The FA group was the reference. The values are given as the mean, calculated on the basis of marginal means, with the 95% CI in parentheses§For Neer and OTA/AO classifications, we adjusted for baseline BMI, quality of reduction, and proximal humeral osteoporosis. For participant-reported osteoporosis and proximal humeral osteoporosis, we adjusted for baseline BMI, quality of reduction, and Neer classification. For the remaining subgroup variables, we adjusted for proximal humeral osteoporosis, quality of reduction, and Neer classification Results of the generalized linear models for subgroup analysis in 12-month DASH score *Numbers in parentheses represent the number of participants in this subgroup **indicates statistically significant results (p < 0.05) &These are the observed values given as the mean (SD) †The FA group was the reference. The values are given as the mean, calculated on the basis of marginal means, with the 95% CI in parentheses §For Neer and OTA/AO classifications, we adjusted for baseline BMI, quality of reduction, and proximal humeral osteoporosis. For participant-reported osteoporosis and proximal humeral osteoporosis, we adjusted for baseline BMI, quality of reduction, and Neer classification. For the remaining subgroup variables, we adjusted for proximal humeral osteoporosis, quality of reduction, and Neer classification Surprisingly, we found that the statistically significant mean difference in the two-part Neer fracture classification(unadjusted mean difference (11.8 [95% CI 1.2 to 22.4]; p = 0.03) and the adjusted mean difference [11.0 [95% CI 4.1 to 17.9]; p = 0.00)], type A fracture classification(adjusted mean difference [10.8 [95% CI 3.4 to 18.3]; p = 0.00)] and diabetes (12.3 [95% CI 0.66 to 23.9]; p = 0.04) subgroups seemed to indicate that using LP alone could have better clinical results in these subgroups. Results validation We selected the proximal humeral osteoporosis category for further analysis as the participants with proximal humeral osteoporosis treated with FA showed better clinical outcomes in DASH scores at 12 months than those who treated with LP alone. Although the participants with alcohol drinking also showed significant difference in the unadjusted results, there was no significant difference in the adjusted results. This indicated that the significant result might be caused by the mismatch of other important factors(proximal humeral osteoporosis, quality of reduction, Neer classification, and so on) between the two groups. We did not select alcohol drinking category for further analysis. Participants in four groups (nonproximal humeral osteoporosis LP group, nonproximal humeral osteoporosis FA group, proximal humeral osteoporosis LP group, and proximal humeral osteoporosis FA group) all reported reduced disabilities over the 12 months after the surgical procedure. The DASH scores at 3, 6, and 12 months are presented in Table 3 and Fig. 3. The unadjusted and adjusted mean differences (−11.3 [95% CI −22.8 to 0.27]; −10.0 [95% CI −21.4 to 1.4]) of DASH score at 6 months were greater than MCID and showed similar trend toward favoring FA augmentation of participants with proximal humeral osteoporosis though it was not statistically significant (p > 0.05). Table 3DASH score of proximal humeral osteoporosis subgroup at 3, 6, and 12 monthsTime (month)Proximal humeral osteoporosisFA guoup dash*LP group dash*Difference†p valueAdjusted difference†§Adjusted p value3Without32.1(2.3)30.3(2.3)1.8(−4.6 to 8.2)0.571.7(−4.7 to 8.1)0.60With30.1(3.1)36.1(3.0)−6.1(−14.4 to 2.2)0.15−5.5(−13.6 to 2.6)0.186Without25.5(2.2)25.5(2.2)0.10(−6.1 to 6.2)0.990.1(−5.9 to 6.1)0.97With19.2(4.3)30.5(4.1)−11.3(−22.8 to 0.27)0.05−10.0(−21.4 to 1.4)0.0912Without18.3(2.3)13.2(2.3)5.1(−1.2 to 11.5)0.115.7(−0.3 to 11.7)0.06With18.4(3.5)31.1(3.4)−12.8(−22.2 to −3.2)0.01−11.1(−20.1 to −2.2)0.01*These are the observed values given as the mean (SD)†The FA group was the reference. The values are given as the mean, calculated on the basis of marginal means, with the 95% CI in parentheses§These values were adjusted for baseline BMI, quality of reduction, and Neer classificationFig. 3Postoperative DASH scores over 12 months for proximal humeral osteoporosis category Table 3DASH score of proximal humeral osteoporosis subgroup at 3, 6, and 12 monthsTime (month)Proximal humeral osteoporosisFA guoup dash*LP group dash*Difference†p valueAdjusted difference†§Adjusted p value3Without32.1(2.3)30.3(2.3)1.8(−4.6 to 8.2)0.571.7(−4.7 to 8.1)0.60With30.1(3.1)36.1(3.0)−6.1(−14.4 to 2.2)0.15−5.5(−13.6 to 2.6)0.186Without25.5(2.2)25.5(2.2)0.10(−6.1 to 6.2)0.990.1(−5.9 to 6.1)0.97With19.2(4.3)30.5(4.1)−11.3(−22.8 to 0.27)0.05−10.0(−21.4 to 1.4)0.0912Without18.3(2.3)13.2(2.3)5.1(−1.2 to 11.5)0.115.7(−0.3 to 11.7)0.06With18.4(3.5)31.1(3.4)−12.8(−22.2 to −3.2)0.01−11.1(−20.1 to −2.2)0.01*These are the observed values given as the mean (SD)†The FA group was the reference. The values are given as the mean, calculated on the basis of marginal means, with the 95% CI in parentheses§These values were adjusted for baseline BMI, quality of reduction, and Neer classification DASH score of proximal humeral osteoporosis subgroup at 3, 6, and 12 months *These are the observed values given as the mean (SD) †The FA group was the reference. The values are given as the mean, calculated on the basis of marginal means, with the 95% CI in parentheses §These values were adjusted for baseline BMI, quality of reduction, and Neer classification Postoperative DASH scores over 12 months for proximal humeral osteoporosis category For the radiographic outcomes, the changes in NSA and HHH are presented in Table 4, Figs. 4 and 5. In four groups, the humeral head height remained stable and the neck-shaft angle decreased slightly which presented a trend of varus alignment over time. There was no significant difference in radiographic outcomes between the FA group and the LP group in participants with or without proximal humeral osteoporosis. Table 4Radiographic outcomes of proximal humeral osteoporosis category at 3, 6, and 12 monthsTime (month)Proximal humeral osteoporosisChanges in NSA of FA group*Changes in NSA of LP group*Difference†p value1Without−0.84(1.92)−3.99(1.88)3.16(−2.12 to 8.43)0.24With−4.81(2.51)−0.53(2.41)−4.28(−11.1 to 2.55)0.223Without−2.30(2.15)−5.59(2.10)3.28(−2.60 to 9.18)0.28With−8.01(2.69)−4.72(2.58)−3.30(−10.6 to 4.10)0.386Without−3.78(2.07)−6.60(2.03)2.82(−2.87 to 8.51)0.33With−5.90 (3.35)−2.66(3.22)−3.24(−12.34 to 5.86)0.4912Without−6.49(2.25)−6.61(2.21)0.12(−6.06 to 6.30)0.97With−9.18(2.97)−5.36(2.85)−3.81(−11.89 to 4.26)0.35Time (month)Proximal humeral osteoporosisChanges in HHH of FA group*Changes in HHH of LP group*Difference†p value1Without0.23(0.75)−1.12(0.74)1.35(−0.71 to 3.42)0.20With−1.01(1.11)−0.10(1.07)−0.91(−3.93 to 2.11)0.563Without−0.56(0.76)0.14(0.75)−0.69(−2.78 to 1.40)0.52With−0.67(0.93)0.11(0.90)−0.79(−3.33 to 1.75)0.546Without−0.29(0.80)0.25(0.78)−0.53(−2.73 to 1.40)0.52With−0.08(0.99)−0.35(0.95)0.27(−2.42 to 2.96)0.8512Without−1.00(0.78)−0.54(0.77)−0.47(−2.63 to 1.70)0.67With−1.17(0.92)−0.21(0.88)−0.96(−3.46 to 1.54)0.45*These are the observed values given as the mean (SD)†The FA group was the reference. The values are given as the mean, calculated on the basis of marginal means, with the 95% CI in parenthesesFig. 4Postoperative changes in head height of humerus (HHH) over 12 months for proximal humeral osteoporosis categoryFig. 5Postoperative changes in neck shaft angle (NSA) over 12 months for proximal humeral osteoporosis category Table 4Radiographic outcomes of proximal humeral osteoporosis category at 3, 6, and 12 monthsTime (month)Proximal humeral osteoporosisChanges in NSA of FA group*Changes in NSA of LP group*Difference†p value1Without−0.84(1.92)−3.99(1.88)3.16(−2.12 to 8.43)0.24With−4.81(2.51)−0.53(2.41)−4.28(−11.1 to 2.55)0.223Without−2.30(2.15)−5.59(2.10)3.28(−2.60 to 9.18)0.28With−8.01(2.69)−4.72(2.58)−3.30(−10.6 to 4.10)0.386Without−3.78(2.07)−6.60(2.03)2.82(−2.87 to 8.51)0.33With−5.90 (3.35)−2.66(3.22)−3.24(−12.34 to 5.86)0.4912Without−6.49(2.25)−6.61(2.21)0.12(−6.06 to 6.30)0.97With−9.18(2.97)−5.36(2.85)−3.81(−11.89 to 4.26)0.35Time (month)Proximal humeral osteoporosisChanges in HHH of FA group*Changes in HHH of LP group*Difference†p value1Without0.23(0.75)−1.12(0.74)1.35(−0.71 to 3.42)0.20With−1.01(1.11)−0.10(1.07)−0.91(−3.93 to 2.11)0.563Without−0.56(0.76)0.14(0.75)−0.69(−2.78 to 1.40)0.52With−0.67(0.93)0.11(0.90)−0.79(−3.33 to 1.75)0.546Without−0.29(0.80)0.25(0.78)−0.53(−2.73 to 1.40)0.52With−0.08(0.99)−0.35(0.95)0.27(−2.42 to 2.96)0.8512Without−1.00(0.78)−0.54(0.77)−0.47(−2.63 to 1.70)0.67With−1.17(0.92)−0.21(0.88)−0.96(−3.46 to 1.54)0.45*These are the observed values given as the mean (SD)†The FA group was the reference. The values are given as the mean, calculated on the basis of marginal means, with the 95% CI in parentheses Radiographic outcomes of proximal humeral osteoporosis category at 3, 6, and 12 months *These are the observed values given as the mean (SD) †The FA group was the reference. The values are given as the mean, calculated on the basis of marginal means, with the 95% CI in parentheses Postoperative changes in head height of humerus (HHH) over 12 months for proximal humeral osteoporosis category Postoperative changes in neck shaft angle (NSA) over 12 months for proximal humeral osteoporosis category We also used the participant-reported osteoporosis variable for results validation, with the findings presented in Appendix Tables 7, 8 and Appendix Figs. 6, 7, 8. The unadjusted and adjusted mean differences (−11.0 [95 CI −26.4 to 4.3]); −21.2 [95% CI −44.1 to 1.8]) of DASH score at 6 months were greater than MCID though it was not statistically significant (p > 0.05). There was no significant difference in radiographic outcomes between the FA group and the LP group in participants with or without participant-reported osteoporosis. Discussion In this subgroup analysis of the previous RCT, we found that for adults with osteoporotic proximal humeral fractures in combination with medial column comminution, FA augmentation + LP fixation yielded statistically and clinically better outcomes of DASH score at 12 months than using an LP alone. The DASH scores at other time points showed the similar trends though did not reach statistical significance. Previous research, in addition to our study, supports the benefits of fibular allografts in treating osteoporotic proximal humeral fractures. A case report from South Korea described that an 88 year old woman with severe osteoporosis and comminuted proximal humeral fracture was successfully treated with fibular allograft augmentation and achieved good therapeutic effects during one-year follow-up [20]. A comparative study showed that in osteoporotic participants the FA group had better forward elevation function and fewer postoperative complications than using an LP alone [16]. Kim DS et al. suggested that fibular allografts augmentation yields better clinical outcomes than using additional inferomedial screws (IMS) for elderly participants with Neer type 3 and 4 proximal humeral fractures, due to the weaker support provided by IMS for osteoporotic humeral heads [21]. In combination with the findings of this study, for participants with osteoporosis of the proximal humerus, the medial support provided by a locking plate alone seems to be relatively insufficient; the insertion of a FA could offer additional and effective medial support in such cases. In addition to the proximal humeral osteoporosis category, we also found significant differences in the diabetes and fracture classification categories. There was a statistically significant difference in the adjusted mean difference of DASH scores between the two groups of participants with diabetes, but should be aware of the fact that the number of patients was very small (only four patients with diabetes were treated with FA). Moreover, in the Neer two-part fractures, LP alone presented better DASH scores than adding FA augmentation. The reason might be that we allowed surgeons to implant cancellous bone grafts in the LP group, which might had improved fixation stability to some extent. Another reason could be that the use of fibular allograft in two-part fractures might damage the soft tissues, thereby affecting postoperative function. Based on the results of this subgroup analysis and the aforementioned studies [16, 20, 21], we believe that despite its potential disadvantages (higher costs, greater surgical complexity, and so on), fibular allograft augmentation can be beneficial in the treatment of medial column comminuted proximal humeral fractures in patients with osteoporosis. In interpreting the results, we place primary emphasis on the effect sizes (mean differences in DASH scores) and their 95% confidence intervals rather than on p values alone. In the proximal humeral osteoporosis subgroup, the adjusted mean difference of −11.1 points (95% CI −20.0 to −2.2) reached statistical significance and exceeded the MCID of ten points, suggesting a potentially clinically relevant benefit of fibular allograft augmentation. However, statistical significance should not be conflated with clinical relevance. The relatively wide confidence intervals indicate limited precision in the effect size estimates, reflecting the modest subgroup sample sizes and the inherent variability of patient-reported DASH scores. Consequently, the true treatment effect could plausibly range from a small difference of uncertain clinical importance to a more substantial improvement. These considerations, together with the exploratory and post hoc nature of the subgroup analysis (not prespecified in the original trial and performed without stratified randomization), necessitate a cautious approach to inferring clinical implications. Although multivariable adjustment was performed for selected covariates (e.g., BMI, Neer classification, and quality of reduction), residual confounding and bias cannot be fully eliminated in this post hoc setting. Notably, this study did not include parameters such as blood loss and operation duration in evaluating surgical outcomes, potentially limiting the comprehensiveness of our assessment. Future research should aim to include these and other relevant factors in order to provide a more comprehensive assessment. We did not find any statistically significant differences in the radiographic outcomes among all the subgroups (including the osteoporosis subgroups), which seems to imply differences in subjective shoulder function benefits could not be explained by the imaging findings. The improved DASH scores in the osteoporosis group following FA augmentation may be attributed to factors beyond the measured NSA and HHH. For example, the FA may provide a scaffold for new bone formation, partially compensating for impaired bone healing in patients with osteoporosis. However, these biological advantages may not be reflected on the measured radiographic outcomes. Besides, the improved DASH scores could also be related to quality of reduction, operation time, comorbidities as well as early mobilization. Moreover, we allowed surgeons to implant cancellous bone grafts in the LP group (mainly used to discover cavities below large nodules), which might had improved fixation stability to some extent [22]. Our findings highlight the complexity of functional recovery following proximal humeral fractures, emphasizing the need for a multifaceted approach to treatment evaluation. Future studies incorporating advanced imaging techniques, biomechanical assessments, and histological analysis of graft integration may provide further insight into the mechanisms underlying the functional benefits observed in our study. This study had limitations. First, the sample size for subgroup analysis might be insufficient. All the nonsignificant results could be caused by insufficient sample size and should be interpreted with caution. The DASH score trajectory of the proximal humeral osteoporosis FA group consistently remained below that of the proximal humeral osteoporosis LP group. This suggested that we might be able to obtain significant results at all the time points if the sample size were sufficient. Second, the subgroup analysis was not prespecified; randomization was not performed stratified by selected subgroups. This inevitably resulted in demographic differences between the two interventions in subgroups; thus, we adjusted the analyses for potential confounders. Furthermore, since the subgroup analyses were not prespecified, all subgroup findings should be regarded as hypothesis-generating rather than confirmatory and are susceptible to an elevated risk of type I error. Third, although we measured radiological outcomes (including Tingart index, NSA, and HHH) through two reviewers, the intra-observer measurement error could not be ignored. Finally, because our results were not adjusted for multiple comparisons, the likelihood of type I error is substantially increased. Although interaction tests were conducted to identify potential heterogeneity of treatment effect, these tests themselves contribute to the overall multiplicity burden. Therefore, all subgroup findings should be interpreted with caution. Conclusions LP + FA augmentation seems to offer superior postoperative clinical outcomes, as measured by DASH scores, compared to LP alone in patients with osteoporotic proximal humeral fractures and medial column comminution. This finding is preliminary and limited by a small sample size, post hoc subgroup design, and potential intra-observer measurement variability. Confirmation of the results requires further investigation through prospective studies with a large sample size, at best by RCTs.