Comparison of the predictive performance of Chang and Baumgaertner reduction quality criteria for cutout complication in geriatric unstable intertrochanteric femur fractures Authors: Yilmaz Onder, Tugrul Bulut, Tolgacan Kurtulus, Mert Bahadirli, Mahmut Tuncez Published: 2026 ABSTRACT AbstractIntroductionThe aim of this study was to compare the predictive performance of Baumgaertner Reduction Quality Criteria (BRQC) and Chang Reduction Quality Criteria (CRQC) for lag screw cutout in geriatric unstable intertrochanteric femur fractures treated with proximal femoral nailing (PFN).Materials and methodsA retrospective study was conducted on 907 patients aged ≥ 65 years who underwent surgery for AO/OTA 31.A2 fractures between January 2018 and January 2026. Among these, 108 patients treated with the same implant (APFN) and meeting the inclusion criteria were included. To reduce the effect of known implant-position–related risk factors for cutout and thereby better isolate the predictive value of reduction quality classifications, patients with postoperative TAD values ≤ 30 mm and ≥ 20 mm and lag screws positioned center-center zone according to Cleveland were included in the study. Reduction quality was evaluated according to BRQC and CRQC using anteroposterior and lateral radiographs. Inter-system agreement was assessed using the McNemar–Bowker test and Cohen’s kappa. Cutout predictive performances were compared using ROC analysis and the DeLong test. Inter- and intra-observer reliability was evaluated using weighted Cohen’s kappa coefficients. A p-value < 0.05 was considered statistically significant.ResultsThe mean age of the patients was 79 ± 7.5 years, and 69.4% were female. 93 patients (86.1%) had bone union (non-cutout group), and 15 patients (13.9%) had cutout (cutout group). The mean bone union time was 12.54 ± 2.65 weeks. The mean time to cutout was 8.29 ± 5.52 weeks. No demographic differences were observed between groups. Both BRQC and CRQC significantly predicted cutout. However, cross-distribution and agreement analyses showed that reduction types associated with cutout were categorized into different risk groups, with CRQC demonstrating greater selectivity. ROC analysis showed AUC values of 0.79 (95% CI: 0.69–0.90) for CRQC and 0.68 (95% CI: 0.56–0.80) for BRQC, with a significant difference (ΔAUC = 0.10, p < .05). The reliability results demonstrated that CRQC exhibited consistently higher inter- and intra-observer agreement compared to BRQC.ConclusionCRQC more accurately and selectively identifies reduction types associated with cutout risk compared to BRQC. The use of CRQC can contribute to more effective intraoperative decision-making and postoperative risk stratification, especially in geriatric unstable intertrochanteric femur fractures.Keywords: Geriatric, Unstable intertrochanteric femur fracture, Reduction quality, Chang and Baumgaertner reduction quality criteria, Cutout Introduction Reduction quality is one of the most important factors determining clinical outcomes after surgical fixation of geriatric unstable intertrochanteric femur fractures [1–3]. Defining and measuring the quality of reduction is of particular importance both for understanding complications and for developing preventive strategies. Although Baumgaertner reduction quality criteria (BRQC) continue to be widely used [4–6], Chang reduction quality criteria (CRQC) has emerged as an alternative assessment method in evaluating reduction quality by defining the concept of anteromedial cortical support (AMCS) [7, 8]. This concept is particularly important for geriatric unstable intertrochanteric femur fractures, because in these patients the posteromedial cortex is usually fragmented. Both BRQC and CRQC classify reduction quality into three categories (poor, acceptable, good/excellent). Theoretically, the risk of cutout complication is expected to increase as the reduction quality moves from “good/excellent” to “poor”. Both methods are widely used in the literature [5–8]. However, it remains unclear which method performs better. Recent studies have continued to refine these methods or propose new scoring systems to improve prediction of cutout risk [9–11]. Determining which method provides better predictive performance is important both scientifically and for routine clinical practice. To our knowledge, there is no study that directly compares the predictive value of BRQC and CRQC in a standardized, homogenous cohort to determine which method is more informative in predicting cutout complications in geriatric unstable intertrochanteric femur fractures treated with PFN. These data are crucial for determining which method is more appropriate and reliable. The hypothesis of this study is that in geriatric unstable intertrochanteric femur fractures, the posteromedial cortex is usually fragmented, making anteromedial cortex continuity more critical, and therefore, CRQC, which also considers anteromedial cortex continuity, is more useful comparing to BRQC in predicting cutout complications. The aim of this study is to compare the predictive performance of CRQC and BRQC in predicting lag screw cutout development in geriatric unstable intertrochanteric femur fractures by standardizing patient and fixation-related variables, and to determine which reduction assessment system predicts cutout risk more accurately and reliably. Materials and methods All procedures involving human participants were performed in accordance with the ethical standards of the institutional committee. This retrospective study was approved by the Institutional Review Board of our university (IRB#14/2026). Patient inclusion A total of 907 patients aged ≥ 65 years who underwent surgical treatment for AO/OTA 31.A2 fracture between January 2018 and January 2026 were evaluated. This was a retrospective registry-based review with no prospective data collection. Inclusion criteria Patients aged ≥ 65 years with AO/OTA 31-A2 intertrochanteric femur fractures who were treated with a single APFN implant (APFN® Antirotator Proximal Femoral Nail, TST Medical Devices Inc., Istanbul, Turkey), had a tip–apex distance (TAD) between 20 and 30 mm (20 ≤ TAD ≤ 30 mm) on early postoperative radiographs, and had the lag screw positioned within the Cleveland–Bosworth centre–centre zone were included. Only patients with adequate radiographic data and a minimum follow-up of 6 months were included. Exclusion criteria Patients with fracture patterns other than AO/OTA 31-A2, pathological fractures, periprosthetic fractures, or fractures resulting from high-energy trauma were excluded. Additional exclusion criteria were age < 65 years, treatment with non-PFN methods or implants other than APFN, Singh index < 3, Charlson comorbidity index (CCI) ≥ 5, follow-up < 6 months, inadequate radiological data, history of fracture or surgery in the ipsilateral or contralateral hip, cumulative ambulatory score (CAS) ≤ 4, being completely bedridden or fully dependent prior to fracture, TAD > 30 mm or < 20 mm, and lag screw placement outside the Cleveland–Bosworth centre–centre zone. Lag screw position was assessed using the Cleveland-Bosworth nine-zone classification. Only screws located in the centre-centre position (zone 5), defined as central placement on both anteroposterior and lateral radiographs, were included. A TAD range of 20–30 mm was deliberately selected based on evidence suggesting that TAD values < 20 mm in PFNA systems may be associated with failure mechanisms such as cut-through (medial migration) [3, 12], and whereas values above approximately 30 mm have been strongly associated with an increased risk of cutout [13]. By excluding these extreme TAD values, we aimed to reduce implant-related confounding and allow a more reliable assessment of the independent effect of reduction quality. In this way, a homogeneous cohort was obtained by standardizing known risk factors, including implant position related to cutout development, in a group of geriatric patients with unstable intertrochanteric femoral fractures aged ≥ 65 years who had the same fracture type and were treated with the same implant. The aim was to more clearly demonstrate the independent predictive power of CRQC and BRQC for cutout development. A total of 108 patients were included in the study (Fig. 1). The diagram illustrates the flow of participants through each stage of the study All procedures were performed by experienced orthopaedic surgeons at a single institution using standard cephalomedullary nailing techniques. Postoperative management consisted of early mobilization without formal weight-bearing restrictions. Patients were mobilized with weight-bearing as tolerated using assistive walking devices under the supervision of physiotherapists. Evaluation of patients Detailed demographic and clinical data were recorded for all patients. Radiological assessments were performed digitally using the picture archiving and communication system (PACS). Radiological evaluations of the patients were performed using preoperative and early postoperative plain anteroposterior (AP) and lateral radiographs. First, all 108 cases were evaluated by two independent observers-one orthopedic specialist and one radiologist-who were not involved in the initial planning or surgery to prevent potential bias, and decisions were reached by consensus. Second, to assess the inter- and intra-observer reliability of the BRQC and CRQC classifications, a subset (30 randomly selected cases) was independently evaluated by two different observers (orthopaedic specialist) who did not participate in the initial assessment, and re-evaluated ten days later. Radiological evaluation The AO/OTA fracture type was recorded for all patients, and the Singh index was evaluated preoperatively from contralateral proximal femur radiographs by the same observers to ensure consistency [14, 15]. The position of the lag screw in the femoral head was determined using TAD and Parker’s ratio method [16, 17]. The placement of the screw in the femoral head was evaluated by dividing it into zones using the Cleveland-Bosworth method [18]. The early postoperative reduction quality was classified using BRQC and CRQC [5–8]. According to BRQC, reduction quality was evaluated using alignment and displacement parameters. (1) Alignment parameters; was defined as a normal or slightly valgus neck-shaft angle on the AP radiograph, and < 20° angulation on the lateral radiograph. (2) Displacement parameters; was defined as < 4 mm displacement in any fragment on both AP and lateral radiographs. According to the BRQC, meeting both criteria was considered “good reduction”, meeting only one was considered “acceptable reduction”, and meeting neither was considered “poor reduction” (Fig. 2). Baumgaertner reduction quality criteria (BRQC) According to CRQC, reduction quality was evaluated using a scoring method based on 4 sub-criteria over alignment and displacement parameters. These criteria were defined as: 1) Normal or slightly valgus neck-shaft angle on AP radiograph, 2) < 20° angulation on lateral radiograph, 3) Neutral or positive medial cortical support on AP radiograph, 4) Smooth contact of the anterior cortexes (neutral or positive anterior cortical support) on lateral radiograph. Each met criterion was recorded as 1 point (total 0–4 points), and based on the total score, 4 points were classified as “excellent reduction”, 2–3 points as “acceptable reduction”, and 0–1 point as “poor reduction” (Fig. 3). Chang reduction quality criteria (CRQC) Cutout diagnosis was made in patients without signs/history of postoperative period trauma or infection, after comparison with previous control plain radiographs. Newly developed lateral migration of the lag screw, the perforation of the lag screw through the femoral neck or head, varus collapse in the head-neck angle and/or rotational deformity in the head-neck segment were considered as cutout [19] (Fig. 4). A patient who developed a cutout complication at postoperative 8th week. In this case, the reduction was classified as “acceptable” according to CRQC and “good” according to BRQC. The occurrence of cutout in this patient suggests that CRQC may provide a more cautious and risk-sensitive classification approach Radiological bone union was defined as the observation of cortical bone bridge formation in both AP and lateral plain radiographs [20]. Statistical analysis Statistical analyses were performed using IBM SPSS Statistics for Windows software (Version 26.0; IBM Corp., Armonk, NY, USA). Descriptive statistics are presented as mean ± standard deviation and median (minimum–maximum) for continuous variables, and frequency and percentage for categorical variables. The distribution of continuous variables was assessed using the Kolmogorov–Smirnov test together with graphical methods, and the homogeneity of variances was evaluated using Levene’s test. For comparisons between two independent groups, the independent samples t-test was applied for normally distributed variables, whereas the Mann–Whitney U test was used for non-normally distributed variables. Categorical variables were compared using the Pearson chi-square test or Fisher’s exact test, depending on expected cell frequencies. Since BRQC and CRQC are three-category (poor/acceptable/good-excellent) paired nominal variables applied to the same cases, the McNemar–Bowker test was used to evaluate the difference between these two classification systems. Analyses were performed separately for cutout group, non-cutout group, and stratified across the total sample (all patients). The level of agreement between classification systems was assessed using Cohen’s kappa coefficient and interpreted according to the Landis–Koch classification. However, since the primary clinical objective in terms of cutout development is the ability to distinguish between reduction types leading to cutout/failure rather than inter-system concordance, the predictive performance of BRQC and CRQC was also examined using receiver operating characteristic (ROC) analysis. Reduction quality was entered into ROC analyses as an ordinal predictor with three ordered categories (1 = poor, 2 = acceptable, 3 = good/excellent). Higher scores represented better reduction quality and, therefore, a lower expected probability of cutout. To ensure that higher test values corresponded to higher predicted event risk in the ROC procedure, the variable was analysed using the software option that specifies the event direction (cutout = positive state), rather than manually reversing or recoding the predictor values. This approach preserved the natural ordering of categories while allowing correct estimation and interpretation of AUC values. In the ROC analysis, the area under the curve (AUC) and 95% confidence intervals were calculated; the AUC values of the two classification systems were compared using the DeLong method. The visual representation of ROC curves was created using the Matplotlib library with the Python programming language (version 3.10). In statistical tests, a two-sided p-value of < 0.05 was accepted as the statistical significance threshold. A post-hoc power analysis was additionally performed for the primary comparison of correlated ROC curves (CRQC vs. BRQC) based on the observed AUC difference, sample size, and event prevalence. Power estimation was conducted using the Hanley–McNeil framework implemented in R statistical software (pROC package). In addition, the minimum detectable AUC difference at 80% power was estimated under the observed sample structure. In addition, a sensitivity analysis was performed using the original CRQC point score (0–4) as a continuous predictor. The inter- and intra-observer agreement was quantified using weighted Cohen’s kappa coefficients appropriate for ordinal data. Both linear and quadratic weighting schemes were applied. 95% confidence intervals (95% CI) were estimated using nonparametric bootstrap resampling (5,000 iterations). Kappa values were interpreted according to the Landis–Koch classification as follows: <0.20 poor, 0.21–0.40 fair, 0.41–0.60 moderate, 0.61–0.80 substantial, and > 0.80 almost perfect agreement. Results The mean age of the 108 patients included in the study was 79 ± 7.5 years, the median age was 79 (65–95) years, the mean CCI value was 3.32 ± 0.93 and the median CCI value was 4 (3–4). 33 (30.6%) of the patients were male, and 75 (69.4%) were female. During the follow-up period, bone union (non-cutout) was achieved in 93 patients (86.1%), while cutout complications developed in 15 patients (13.9%). The mean bone union time was recorded as 12.54 ± 2.65 weeks, and the mean time to cutout as 8.29 ± 5.52 weeks. Reduction quality was classified as “good” in 68 (63.0%) patients by BRQC, whereas CRQC classified 52 (48.1%) patients as having “excellent” reduction (Table 1). Of the 15 patients who developed cutout, 14 were treated with hemiarthroplasty as a salvage procedure, while one patient declined further surgical intervention. Table 1Descriptive characteristics of patientsDemographic variablesDataAge (year) *79 ± 7.579 (65–95)Sex n (%) Male Female33 (30.6)75 (69.4)CCI *3.32 ± 0.934 (3–4)Group n (%) Non-cutout / Union Cutout93 (86.1)15 (13.9)Cutout time (week) *8.29 ± 5.524.7 (1–24)Union time (week) *12.54 ± 2.6514 (6–24)BRQC n (%) Poor Acceptable Good2 (1.9)38 (35.1)68 (63.0)CRQC n (%) Poor Acceptable Excellent11 (10.2)45 (41.7)52 (48.1)n: Number of patients, %: Column percentageCCI Charlson comorbidity index, BRQC Baumgaertner reduction quality criteria, CRQC Chang reduction quality criteria*The data is presented as mean ± standard deviation and median (minimum-maximum) Descriptive characteristics of patients 79 ± 7.5 79 (65–95) Male Female 33 (30.6) 75 (69.4) 3.32 ± 0.93 4 (3–4) Non-cutout / Union Cutout 93 (86.1) 15 (13.9) 8.29 ± 5.52 4.7 (1–24) 12.54 ± 2.65 14 (6–24) Poor Acceptable Good 2 (1.9) 38 (35.1) 68 (63.0) Poor Acceptable Excellent 11 (10.2) 45 (41.7) 52 (48.1) n: Number of patients, %: Column percentage CCI Charlson comorbidity index, BRQC Baumgaertner reduction quality criteria, CRQC Chang reduction quality criteria *The data is presented as mean ± standard deviation and median (minimum-maximum) When the defining characteristics were compared between groups, no significant differences were found in terms of age, gender and CCI value (for each p>.05). However, in both BRQC and CRQC, significantly lower cutout rates were found in patients in the “good and excellent reduction” categories. In the cutout group, according to BRQC, “good reduction” decreased while “acceptable reduction” increased, and according to CRQC, “excellent reduction” decreased while “poor reduction” increased, and these changes were statistically significant. Both methods significantly predicted cutout occurrence (for each p<.05) (Table 2). Table 2Demographic variables of groupsVariablesNon-cutout (n = 93)Cutout (n = 15)p-valueAge (year) *79.1 ± 7.679 (65–95)78.3 ± 6.978 (65–91)0.704†Sex n (%) Male Female31 (33.3)62 (66.7)2 (13.3)13 (86.7)0.208&CCI *3.30 ± 0.923.40 ± 0.990.900†3 (3–4)4 (3–4)BRQC n (%) Poor Acceptable # Good #2 (2.2) a26 (28) a65 (69.8) a0 (0.0) a12 (80) b3 (20) b < 0.001 ƞ CRQC n (%) Poor # Acceptable Excellent #4 (4.3) a38 (40.9) a51 (54.8) a7 (46.7) b7 (46.7) a1 (6.6) b < 0.001 ƞ n: Number of patients, %: Column percentageCCI Charlson comorbidity index, BRQC Baumgaertner reduction quality criteria, CRQC Chang reduction quality criteria*The data is presented as mean ± standard deviation and median (minimum-maximum), #The variables that cause the statistically significant difference. If both columns contain “a”, there is no difference; if one column contains “a” and the other contains “b”, then there is a difference, †Mann-Whitney U test, &: Chi-square analysis, ƞ: Fisher exact testp-values in bold indicate statistical significance Demographic variables of groups 79.1 ± 7.6 79 (65–95) 78.3 ± 6.9 78 (65–91) Male Female 31 (33.3) 62 (66.7) 2 (13.3) 13 (86.7) Poor Acceptable # Good # 2 (2.2) a 26 (28) a 65 (69.8) a 0 (0.0) a 12 (80) b 3 (20) b Poor # Acceptable Excellent # 4 (4.3) a 38 (40.9) a 51 (54.8) a 7 (46.7) b 7 (46.7) a 1 (6.6) b n: Number of patients, %: Column percentage CCI Charlson comorbidity index, BRQC Baumgaertner reduction quality criteria, CRQC Chang reduction quality criteria *The data is presented as mean ± standard deviation and median (minimum-maximum), #The variables that cause the statistically significant difference. If both columns contain “a”, there is no difference; if one column contains “a” and the other contains “b”, then there is a difference, †Mann-Whitney U test, &: Chi-square analysis, ƞ: Fisher exact test p-values in bold indicate statistical significance The cross-distribution of reduction quality classifications of patients according to BRQC and CRQC was evaluated separately for the total sample (all patients) and for the cutout and non-cutout groups (Table 3). Accordingly, in all patients and in the non-cutout group, the McNemar–Bowker test rejected the assumption of marginal homogeneity between the BRQC and CRQC classifications, and showed a statistically significant difference between the two classification systems (McNemar–Bowker test results respectively; χ²(4) = 62.83, p<.001, χ²(4) = 57.33, p<.001). This result indicates that BRQC and CRQC classify reduction quality similarly, but not identically. In this assessment, it was particularly noteworthy that a significant portion of the cases rated as “good reduction” according to BRQC shifted to the “acceptable reduction” category according to CRQC. In the cutout group, the McNemar–Bowker test was not statistically significant (McNemar–Bowker χ²(2) = 6.07, p=.093). This finding shows that BRQC and CRQC classifications are not consistent in terms of marginal distributions in patients who develop cutout, but this difference does not reach statistical significance due to the limited sample size. This indicates that the same patient can be evaluated in different risk categories by the two systems after fixation. Table 3Cross-tabulation of BRQC and CRQCAll patients (n = 108)BRQCCRQC – PoorCRQC – AcceptableCRQC – GoodTotalPoor1 (50.0%)1 (50.0%)0 (0.0%)2 (100%)Acceptable9 (23.7%)29 (76.3%)0 (0.0%)38 (100%)Good1 (1.5%)15 (22.1%)52 (76.5%)68 (100%)Total11 (10.2%)45 (41.7%)52 (48.1%)108 (100%)McNemar–Bowker χ²(4) = 62.83, p < .001Cutout Group (n = 15)BRQCCRQC – PoorCRQC – AcceptableCRQC – GoodTotalAcceptable7 (58.3%)5 (41.7%)0 (0.0%)12 (100%)Good0 (0.0%)2 (66.7%)1 (33.3%)3 (100%)Total7 (46.7%)7 (46.7%)1 (6.7%)15 (100%)McNemar–Bowker χ²(2) = 6.07, p = .093Non-cutout Group (n = 93)BRQCCRQC – PoorCRQC – AcceptableCRQC – GoodTotalPoor1 (50.0%)1 (50.0%)0 (0.0%)2 (100%)Acceptable2 (7.7%)24 (92.3%)0 (0.0%)26 (100%)Good1 (1.5%)13 (20.0%)51 (78.5%)65 (100%)Total4 (4.3%)38 (40.9%)51 (54.8%)93 (100%)McNemar–Bowker χ²(4) = 57.33, p < .001The McNemar–Bowker test evaluates marginal homogeneity in paired multinomial data, testing whether the distribution of classifications differs between the two systems across categories. A significant result indicates that the two classification systems do not classify patients equivalently. Cohen’s kappa coefficient quantifies the level of agreement beyond chance between the two systems, with values interpreted as follows: <0.20 poor, 0.21–0.40 fair, 0.41–0.60 moderate, 0.61–0.80 substantial, and > 0.80 almost perfect agreementBRQC Baumgaertner reduction quality criteria, CRQC Chang reduction quality criteria Cross-tabulation of BRQC and CRQC The McNemar–Bowker test evaluates marginal homogeneity in paired multinomial data, testing whether the distribution of classifications differs between the two systems across categories. A significant result indicates that the two classification systems do not classify patients equivalently. Cohen’s kappa coefficient quantifies the level of agreement beyond chance between the two systems, with values interpreted as follows: <0.20 poor, 0.21–0.40 fair, 0.41–0.60 moderate, 0.61–0.80 substantial, and > 0.80 almost perfect agreement BRQC Baumgaertner reduction quality criteria, CRQC Chang reduction quality criteria The agreement between BRQC and CRQC classifications was evaluated separately using Cohen’s kappa coefficient, depending on the presence of cutouts. In the non-cutout group, a good level of agreement was found between the two classification systems (κ ≈ 0.63, p<.001). In contrast, in the cutout group, the agreement between BRQC and CRQC was negligible (κ = 0.02 (95% CI: −0.19 to 0.23, p > .05). This situation reveals that both systems categorize reduction types that develop cutout complication into different risk categories, and that CRQC, in particular, takes a more selective approach in these cases. When the predictive performance of BRQC and CRQC for lag screw cutout development was evaluated using ROC analysis, the AUC value for BRQC was 0.68 (95% CI: 0.56–0.80, p < .05), whereas the AUC value for ordinal CRQC classification was 0.79 (95% CI: 0.69–0.90, p < .001) (Table 4). The DeLong test demonstrated that CRQC had significantly higher discriminatory performance than BRQC (ΔAUC = 0.10, p < .05) (Fig. 5). Table 4ROC analysis for prediction of lag screw cutout across BRQC and CRQC representationsModelAUC95% CIp-valueBRQC ordinal0.680.56–0.80 < 0.05 BRQC score (0–2)0.680.56–0.80 < 0.05 CRQC ordinal0.790.69–0.90 < 0.001 CRQC score (0–4)0.820.72–0.92 < 0.001 BRQC ordinal and BRQC score (0–2) yielded the same AUC because they represent monotonic codings of the same three-level ordered construct. Higher AUC values indicate better discrimination for lag screw cutoutROC Receiver operating characteristics, BRQC Baumgaertner reduction quality criteria, CRQC Chang reduction quality criteria, AUC Area under the curve, CI Confidence intervalp-values in bold indicate statistical significance ROC analysis for prediction of lag screw cutout across BRQC and CRQC representations BRQC ordinal and BRQC score (0–2) yielded the same AUC because they represent monotonic codings of the same three-level ordered construct. Higher AUC values indicate better discrimination for lag screw cutout ROC Receiver operating characteristics, BRQC Baumgaertner reduction quality criteria, CRQC Chang reduction quality criteria, AUC Area under the curve, CI Confidence interval p-values in bold indicate statistical significance Receiver operating characteristics analysis for Baumgaertner and Chang reduction quality criteria In a sensitivity analysis using the original CRQC point score (0–4) as a continuous predictor, the AUC further increased to 0.82 (95% CI: 0.72–0.92, p < .001), supporting the robustness of the primary findings and indicating that preservation of score granularity improves predictive discrimination (Fig. 6). Receiver operating characteristics curves comparing BRQC ordinal classification, BRQC score (0–2), CRQC ordinal classification, and CRQC score (0–4) for prediction of lag screw cutout. The CRQC score demonstrated the highest discriminative performance, followed by ordinal CRQC. BRQC ordinal and BRQC score (0–2) showed overlapping performance The reliability results demonstrated that CRQC exhibited consistently higher inter-observer agreement compared to BRQC in both the first and second evaluations (κw = 0.62 vs. 0.41 and κw = 0.59 vs. 0.38, respectively). Similarly, intra-observer agreement was higher for CRQC than BRQC for both observers (Observer 1: κw = 0.72 vs. 0.58; Observer 2: κw = 0.68 vs. 0.49) (Table 5). Table 5Inter- and intra-observer reliability for BRQC and CRQC classificationsAnalysisCriterionComparisonκ (95% CI)InterpretationInter-observerBRQC1th round0.41 (0.24–0.58)Moderate2nd round0.38 (0.20–0.55)FairCRQC1th round0.62 (0.45–0.78)Substantial2nd round0.59 (0.42–0.75)ModerateIntra-observerBRQCObserver 10.58 (0.40–0.74)ModerateObserver 20.49 (0.31–0.66)ModerateCRQCObserver 10.72 (0.56–0.86)SubstantialObserver 20.68 (0.52–0.83)SubstantialConfidence intervals were estimated using bootstrap resampling. Interpretation based on Landis–Koch criteriaBRQC Baumgaertner reduction quality criteria, CRQC Chang reduction quality criteria, κ Weighted Cohen’s kappa, CI Confidence interval Inter- and intra-observer reliability for BRQC and CRQC classifications Confidence intervals were estimated using bootstrap resampling. Interpretation based on Landis–Koch criteria BRQC Baumgaertner reduction quality criteria, CRQC Chang reduction quality criteria, κ Weighted Cohen’s kappa, CI Confidence interval Discussion The findings of this study suggest that, in geriatric unstable intertrochanteric femur fractures treated with PFN, CRQC provides better discrimination of reduction patterns associated with cutout risk than BRQC. This may be because CRQC captures biomechanically unstable reduction patterns that may be overlooked by classifications based primarily on geometric alignment. Therefore, CRQC may be a more informative assessment tool than BRQC in predicting the development of cutout complication after fixation in geriatric unstable intertrochanteric femur fractures. Although there are many factors that affect stability after fixation in geriatric unstable intertrochanteric femur fractures, reduction quality is one of the most critical parameters in this process [21–23]. Especially in unstable fractures, fragmentation and displacement of the posteromedial cortex negatively affects cortical support and anatomical alignment in this region, decreasing stability. Therefore, restoring and maintaining anteromedial cortical contact is critically important for the continuity of reduction in terms of ensuring load transfer and maintaining mechanical stability [24]. Baumgaertner reduction quality criteria, while also being one of the most frequently used approaches for many years, ignores this highly important biomechanical point. In fact, it has been shown that in unstable fractures, fragmentation often occurs in the posteromedial wall, while the anteromedial wall usually remains more intact and often contains only a single, simple fracture line [25]. In addition, a finite element analysis study showed that 78.7% of the medial wall stability after intertrochanteric fracture is provided by the anteromedial wall and 21.3% by the posteromedial wall [26]. These findings demonstrate that focusing on the anteromedial wall is particularly critical in the reduction of unstable intertrochanteric femoral fractures. CRQC, at this point, differs from BRQC as a relatively newer reduction quality assessment system by defining the concepts of anterior cortical support (ACS) and medial cortical support (MCS). The fundamental difference between BRQC and CRQC stems from their approaches to displacement between fracture fragments. BRQC considers displacements less than 4 mm on AP and lateral radiographs as a good reduction criterion, regardless of the relative position of the fracture fragments. CRQC, on the other hand, focuses on the AMCS concept. In positive MCS, the medial cortex of the proximal fragment is located medial to the medial cortex of the distal fragment, and the displacement is < 1 cortical thickness. In neutral MCS, the alignment of the fragments is correct. In negative MCS, the proximal fragment is located lateral to the distal fragment, regardless of the amount of displacement. In lateral radiographs, excellent reduction is defined by proper contact of the anterior cortices and displacement less than half the cortical thickness. In BRQC, the definition of “good reduction” is based solely on “displacement less than 4 mm,” which allows both positive and negative AMCS to fall within this reduction category. As a result, some patients who could potentially develop a cutout are classified as having “good reduction” according to BRQC, which may obscure the decisive role of reduction quality in the development of cutout complication. Indeed, some cases that developed cutout in present study were classified as “good or acceptable reduction” by BRQC but as “poor reduction” by CRQC, clearly demonstrates the clinical reflection of this conceptual difference. Furthermore, the lack of statistical agreement between BRQC and CRQC classifications in patients who developed cutout in our study indicates that the two systems evaluate reduction types carrying a risk of failure using different criteria. This is not a contradiction, but rather a natural consequence of the differences between the two evaluation systems. Additionally, statistically, the lack of significance in the McNemar–Bowker test and the negligible Cohen’s kappa value in patients developing cutout indicate that the two systems do not perceive the same cases at the same risk level in this critical patient group. Clinically, this finding is extremely important because the aim here is not to achieve consistency between classifications, but to be able to distinguish the reduction types that can cause a cutout complication. In this context, the fact that CRQC tends to have a stricter and more selective classification tendency in patients who develop cutout suggests that it provides a more clinically meaningful risk classification. While BRQC keeps some failed reduction types within “acceptable reduction” limits, CRQC better reflects the true quality of reduction type by placing these cases in categories with worse reduction quality. This suggests that CRQC may identify high-risk reduction patterns more effectively than BRQC. In the cutout group, the discrepancy between classifications suggests that BRQC may be less effective in distinguishing reduction types associated with cutout. ROC analyses conducted to determine the predictive power of evaluation systems for cutout development have shown that CRQC offers higher predictive accuracy than BRQC in evaluating reduction quality. The AUC value of BRQC was calculated as 0.68 (95% CI: 0.56–0.80), showing that it can better distinguish reduction types associated with cutout risk than random estimation; however, its discriminatory power remained moderate. In contrast, CRQC’s AUC was found to be 0.79 (95% CI: 0.69–0.90). This indicates that it exhibits a better discriminative performance in predicting cutout development. A direct comparison using the DeLong test showed that CRQC had a significantly higher AUC value than BRQC (p<.05), supporting the idea that CRQC can more accurately identify reduction types leading to failure. In their study evaluating 127 trochanteric fractures, Mao et al. reported that CRQC is more reliable than BRQC in predicting postoperative mechanical complications [27]. However, the authors did not limit other factors that could affect cutout development and included them in the model as categorical variables. This may be considered an important limitation, as it may reduce the interpretability of the reported predictive performance. Therefore, in our current study, we planned to work on a methodologically homogeneous geriatric patient cohort by standardizing patient-specific factors such as fracture type and bone quality and surgical factors such as TAD, lag screw position/zone and type of PFN, that may affect cutout development, thereby minimizing known potential confounding variables. Thus, the independent effect of reduction quality and the evaluation systems defining it on cutout complications has been more clearly demonstrated. This strengthens the interpretability, internal validity, and generalizability of the findings. In addition, considering the potential impact of observer-dependent variability on the comparative diagnostic performance of BRQC and CRQC, we also conducted a formal reliability analysis to systematically evaluate inter- and intra-observer agreement. Results of this analysis demonstrated that CRQC showed consistently higher inter- and intra-observer agreement compared to BRQC. This is noteworthy, as CRQC incorporates multiple ordinal sub-criteria, which could theoretically increase variability. However, despite its greater structural complexity, CRQC exhibited moderate-to-substantial agreement across all analyses, whereas BRQC showed predominantly moderate agreement. These results are important because they eliminate the possibility that the better predictive performance of CRQC found in ROC analysis could be affected by observer variability. In conclusion, the results of the present study confirmed the decisive role of reduction quality on cutout complication in geriatric unstable intertrochanteric femur fractures, and that CRQC classifies this quality more reliably and effectively. The stronger predictive performance of CRQC compared with BRQC may enable earlier identification of patients at risk of cutout, thereby facilitating timely intraoperative and postoperative interventions. This is particularly valuable in geriatric patients with osteoporotic bone structure. In these patients, the decrease in fixation strength further increases the importance of reduction quality. A classification system that more effectively identifies high-risk reduction types can facilitate the early diagnosis and enable timely interventions (e.g., modifying the intraoperative surgical strategy and restricting postoperative weight-bearing protocols). This study has some limitations. The main limitations are its single-center, retrospective design and relatively limited sample size. Another potential limitation of this study is surgeon-related variability. Although all procedures were performed by experienced orthopaedic surgeons at a single institution following standardized cephalomedullary nailing techniques, minor variations in reduction strategy and intraoperative decision-making may have occurred due to the retrospective nature of the study. An additional potential limitation is the exclusive use of a single cephalomedullary nail system, which may limit the generalisability of the findings. Although this approach improved internal validity by eliminating implant-related variability, the predictive performance of CRQC compared to BRQC may differ with other implant designs. However, the present study also has some distinct strengths. In our study conducted on a homogeneous patient cohort, the use of a single type of implant and the standardization of lag screw position is a conscious methodological choice aimed at largely eliminating known risk factors related to implant for cutout, and thus more clearly revealing the independent predictive value of reduction quality classifications. This approach allowed a clearer assessment of the discriminatory performance of the reduction quality classifications. Conclusion CRQC demonstrated statistically superior AUC in this single-center standardized cohort, supporting its potential value as a biomechanically more comprehensive classification tool; prospective validation in larger, multi-center cohorts with formal reliability assessment is required before recommending its adoption over BRQC in routine clinical practice. Acknowledgements Not applicable. Abbreviations Anterior cortical support Anteromedial cortical support Arbeitsgemeinschaft für Osteosynthesefragen / Orthopaedic Trauma Association Anteroposterior Antirotator proximal femoral nail Area under the curve Baumgaertner reduction quality criteria Calcar-referenced tip–apex distance Cumulative ambulatory score Charlson comorbidity index Chang reduction quality criteria Institutional review board Medial cortical support Picture archiving and communication system Proximal femoral nail Receiver operating characteristic Tip–apex distance Authors’ contributions YO: Conceptualization, methodology, investigation, data collection, formal analysis, writing-original draft, writing-review and editing. TB: Investigation, data collection, writing-review and editing. TK and MB: Data collection, writing-review and editing. MT: Formal analysis, writing-review and editing. Funding This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Data availability The datasets used and/or analysed during the current study are not publicly available due to patient privacy and ethical restrictions but are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate This study was performed in accordance with the principles of the Declaration of Helsinki (as revised in 2013). All procedures were conducted in accordance with institutional ethical standards. The study was approved by the Institutional Review Board of Izmir Katip Celebi University (IRB#14/2026). Due to the retrospective nature of the study, the requirement for informed consent was waived by the ethics committee. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The datasets used and/or analysed during the current study are not publicly available due to patient privacy and ethical restrictions but are available from the corresponding author on reasonable request.