Postoperative Evaluation of Reduction Loss in Proximal Humeral Fractures: A Comparison of Plain Radiographs and Computed Tomography

Abstract

Objective

To compare postoperative CT images with plain radiographs for measuring prognostic factors of reduction loss of fractures of the proximal part of the humerus.

Methods

A total of 65 patients who sustained fractures of the proximal humerus treated with locking plates from June 2012 to October 2015 were retrospectively analyzed. There were 24 men and 41 women, with a mean age of 60.0 years (range, 22–76 years). According to the Neer classification system of proximal humeral fracture, there were 26 two‐part, 27 three‐part and 12 four‐part fractures of the proximal part of the humerus, and all fractures were treated with open reduction and internal fixation (ORIF) using locked plating. All postoperative CT images and plain radiographs of the patients were obtained. Prognostic factors of the reduction loss were the change of neck shaft angle (NSA) and the change of humeral head height (HHH). The change of NSA and HHH were evaluated by the difference between postoperative initial and final follow‐up measurement. Reduction loss was defined as the change ≥10° for NSA or ≥5 mm for HHH. The NSA and HHH were measured using plain radiographs and 3‐D CT images, both initially and at final follow‐up. The paired t‐test was used for comparison of NSA, change of NSA, HHH, and change of HHH between two image modalities. The differences between two image modalities in the assessment of reduction loss were examined using the χ²‐test (McNemar test). Intraclass correlation coefficients (ICC) were used to assess the intra‐observer and inter‐observer reliability.

Results

3‐D CT images (ICC range, 0.834–0.967) were more reliable in all parameters when compared with plain radiographs (ICC range, 0.598–0.915). Significant differences were found between the two image modalities in all parameters (plain radiographs: initial NSA = 133.6° ± 3.8°, final NSA = 130.0° ± 1.9°, initial HHH = 17.9 ± 0.9 mm, final HHH = 15.8 ± 1.5 mm; 3‐D CT: initial NSA = 131.4° ± 3.4°, final NSA = 128.8° ± 1.7°, initial HHH = 16.8 ± 1.2 mm, final HHH = 14.5 ± 1.1 mm; all P < 0.05). In the assessment of reduction loss, the percentage was 16.9% (11/65) for the plain radiographs and 7.7% (5/65) for the 3‐D CT scans (P < 0.05). For the 5 patients with reduction loss, which were observed by two imaging modalities, the mean Constant–Murley score was 61.0 ± 1.6. The patients with reduction loss, observed only in plain radiographs but not CT images, had good shoulder function (Constant–Murley score: 82.7 ± 1.0).

Conclusions

Our data reveal that 3‐D CT images are more reliable than plain radiographs in the assessment of the prognostic factors of reduction loss of fractures of the proximal part of the humerus with treatment of locking plates; this reliable CT technique can serve as an effective guideline for the subsequent clinical management of patients.

Introduction

Proximal humeral fractures account for approximately 5% of all fractures and 45% of fractures of the humerus1, 2. These fractures are the third most common fractures in humans, which mostly affect people older than 65 years of age3. Meanwhile, due to the increase in the prevalence of osteoporotic fractures4, fractures of the proximal humerus are becoming more common, with a 2.5‐fold increase in incidence during the past 30 years5.

Nondisplaced or minimally displaced fractures can be treated non‐operatively, and doing so has achieved good results6. Unstable and displaced fractures of the proximal humerus are commonly treated surgically, including closed reduction and percutaneous pinning, tension band wiring, screw osteosynthesis, intramedullary nails, proximal humerus plates and hemi or reversed arthroplasty. To our knowledge, anatomic reduction and early postoperative functional therapy are essential to the achievement of satisfactory active motion, strength and function7. However, there are no unified conclusions about the optimal management for unstable and displaced proximal humeral fractures8. In recent years, locking plates have been used increasingly for the treatment of proximal humerus fractures, and have achieved good clinical outcomes9, 10, 11, 12. The advantages of the locking compression plates are that they can provide greater angular stability in the proximal humerus than standard plates, and that they are the most rigid under three load tests (axial torque, varus bending, and medial shearing)9, 10, 13. However, remain a significant number of postoperative complications remain, including reduction loss, screw cutout, subacromial impingement, humeral head osteonecrosis, nonunion, loosening, metal failure, and infection11, 12, 13, 14, 15. Reduction loss, occurring in 6.7%–13.7% of cases11, 14, is one of the common and serious complications11, 12, 14, 15. Due to osteoporosis, age and severe comminution11, 12, 14, it is difficult to prevent the reduction loss. Reduction loss can lead to varus malunion and implant failure15, which is the most common reason for reoperations11, 12. Therefore, it is crucial to understand the prognostic factors of the reduction loss.

Previous studies have demonstrated that the postoperative changes in neck shaft angle (NSA) and humeral head height (HHH) are prognostic factors for reduction loss11, 13. Postoperative plain radiographs are routinely utilized to evaluate NSA and HHH13, 16, 17, 18, which are used as important reference values for the subsequent clinical management. Due to the lower costs involved and less radiation exposure, plain radiographs are more acceptable in clinical practice. With use of the postoperative plain radiographs, Jung et al. retrospectively evaluated 252 patients with fractures of the proximal humerus who underwent treatment using locking plates, and found the risk factors for reduction loss11. Gardner et al. evaluated the postoperative change of HHH of 35 patients on the basis of plain radiographs, and stated that mechanical support of the medial region was important for maintenance of reduction for the locking plates of proximal humeral fractures13. With use of similar imaging methods, Zhang et al. assessed the clinical benefit of medial support screws for fractures of the proximal part of the humerus with treatment using locking plates, and found that medial support was of no benefit in Neer two‐part fractures17. However, standard shoulder radiographs are difficult to obtain due to postoperative pain, poor compliance, difference projective angle of the tube, and varying technician experience. Precise measurement of NSA is complicates because of internal and external rotations of the glenohumeral joint16. Therefore, the reliability of measurements, using plain radiographs, is questionable19. Due to the limitations of plain radiographs, it is important to find a new imaging technique to improve the reliability of the evaluation of these prognostic factors.

Over the past 30 years, CT scans have been frequently utilized in orthopaedic and trauma practice. The advantage of 3‐D CT images is that they can provide high quality images for observers, and surgeons can observe the images freely in 3‐D space20. The advantages of 3‐D CT scans are reported on in the context of extra‐articular scapular fractures by Anavian et al. who state that 3‐D CT is more reliable in the assessment of scapula fracture displacement21.

Although CT techniques have many advantages, their value has not been proven in the postoperative assessment of reduction loss of proximal humerus fractures. The purpose of this study was: (i) to present selected techniques to measure the prognostic factors of reduction loss with the use of 3‐D CT scans; (ii) to evaluate the reliability of these measurement techniques in two imaging modalities, including plain radiography and 3‐D CT scans, using interobserver and intraobserver reliability analyses; and (iii) to determine whether there was a difference in the measurements between imaging modalities using comparative statistics. If validated, the standardized measurement method could serve as a reliable technique to evaluate the postoperative factors of proximal humeral fractures.

Materials and Methods

Patients

This study was approved by the local ethics committee, and informed consent was obtained from all patients. We retrospectively evaluated 65 patients who sustained proximal humerus fractures and were treated with proximal humerus internal locking system (PHILOS; Synthes, Stratec Medical, Mezzovico, Switzerland) plates between June 2012 and October 2015. There were 24 men and 41 women, with a mean age of 60.0 years (range, 22–76 years). The inclusion criteria were as follows: (i) age older than 18 years; (ii) closed, unstable fractures of the proximal humeral (two‐part, three‐part, or four‐part fracture according to the Neer classification system)22; and (iii) a follow‐up of at least 6 months and complete postoperative imaging. The exclusion criteria were as follows: (i) pathological fractures, open fractures or stable fractures; and (ii) lacking the necessary imaging date. Based on preoperative radiographs and CT scans, all fractures were classified by the Neer system. There were 26 two‐part, 27 three‐part and 12 four‐part fractures of the proximal humerus.

Radiology Technique

The postoperative plain radiographs and CT images scanned using a 16‐detector spiral CT scanner (GE, Light Speed, Waukesha, WI, USA) were collected. Imaging parameters for CT scanning were as follows: section thickness, 0.625 mm; tube voltage, 120 kVp; pitch, 1.375; matrix, 512 × 512. The thin‐section CT images of all patients were input into the computer‐aided orthopaedics clinical research platform (SuperImage orthopedics edition 1.1, Cybermed, Shanghai, China)23. In this SuperImage system, 3‐D images of the index shoulder joint were generated by surface shaded display (SSD) algorithm with a reconstruction interval of 0.625 mm. Based on the 3‐D interactive and automatic segmentation technique, all bone components were distinguished. The proximal humerus and implant were generated, removing unrelated bones.

Radiographic Evaluation

In plain radiographs, NSA was measured by the intersection of the line vertical to the anatomic neck and the line parallel to the long axis of the humeral shaft, as previously described (Fig. 1A)24. HHH was defined as the vertical distance between two lines, both perpendicular to the shaft of the plate. One was placed at the superior edge of the humeral head, and one was placed at the top edge of the plate (Fig. 1C)13.

Figure 1.

Measurement of neck shaft angle (NSA) and humeral head height (HHH) on plain radiographs and 3‐D CT images. (A) In the plain radiographs, the NSA was measured by the intersection of the line vertical to the anatomic neck and the line parallel to the long axis of the humeral shaft. (B) In the 3‐D CT images, NSA = α + 90°; the head inclination angle (α) was the angle between the humeral shaft axis (line ab) and the anatomic neck plane (plane 1); points a and b were the midpoint of the diameter of the upper humeral shaft. (C) In the plain radiographs, the HHH was defined as the vertical distance between two lines that were both perpendicular to the shaft of the plate. One was placed at the superior edge of the humeral head, and one was placed at the top edge of the plate. (D) In the 3‐D CT images, the HHH was defined as the vertical distance from point d to plane 2; points c and d were the most superior point on the humeral head and the plate, respectively; plane 2 was the plane that was via point c and parallel to the transverse axis of the upper humeral shaft.

In the 3‐D images, the shape of the upper humeral shaft was approximated as a cylinder, as previously reported25. Points a and b were both the midpoint of the diameter of the upper humeral shaft (Fig. 1B). The humeral shaft axis was the line that passed through the above two points. As previously described26, plane 1 was defined as the anatomic neck plane (Fig. 1B). The head inclination angle (α) was the angle between the humeral shaft axis and the anatomic neck plane. NSA was calculated by α plus 90°. Points c and d were the most superior points on the humeral head and the plate, respectively (Fig. 1D). Plane 2 was the plane that was via point c and parallel to the transverse axis of the upper humeral shaft (Fig. 1D). HHH was defined as the vertical distance from point d to plane 2.

The parameters, NSA and HHH, were measured for each radiograph and 3‐D CT image, both initially and at final follow‐up. The change of NSA and HHH was evaluated using the difference between postoperative initial and final follow‐up measurements. Reduction loss was defined as the change ≥10° for NSA or ≥5 mm for HHH11. The measurements were performed by three observers on two separate occasions, 3 weeks apart. These observers included three orthopaedic surgeons with clinical experience of 14, 12, 3 years, respectively. Each observer was blinded to others.

Statistical Analysis

SPSS (version 19.0, Chicago, IL, USA) was utilized for statistical analysis. Reliability of the measurements on plain radiographs and 3‐D CT scans were evaluated using the intraclass correlation coefficient (ICC). An ICC of 0.8 or greater was defined as showing excellent inter‐observer or intra‐observer agreement. The Kolmogorov–Smirnov test was used to examine the normality of all parameters that were found to follow the normal distribution. The data were represented with mean ± standard deviation (SD). The paired t‐test was used for comparison of NSA, change of NSA, HHH, and change of HHH between the two image modalities. The differences between two image modalities in the assessment of reduction loss were examined using the χ²‐test (McNemar test). The level of significance was set at P < 0.05.

Results

Parameters

The measurements for NSA and HHH are summarized in Table 1. Based on plain radiographs, the mean initial NSA and HHH, final NSA and HHH were 133.6° (range, 126.5°–143.3°), 17.9 mm (range, 15.3–19.4 mm), 130.0° (range, 125.2°–134.4°), and 15.8 mm (range, 12.4–18.4 mm), respectively. Based on 3‐D CT images, the mean initial NSA and HHH, final NSA and HHH were 131.4° (range, 126.4°–141.2°), 16.8 mm (range, 14.0–19.2 mm), 128.8° (range, 125.1°–134.5°), and 14.5 mm (range, 12.1–16.5 mm), respectively. Between plain radiographs and CT scans, the aforementioned parameters were significantly different (P < 0.05, Table 1). The mean change of the NSA was 3.6° (range, 0.2°–12.1°) in the plain radiographs and 2.5° (range, 0°–12.3°) in the 3‐D CT images (P < 0.05). The mean change of the HHH was 2.0 mm (range, 0.3–6.5 mm) in the plain radiographs and 2.3 mm (range, 0.4–6.1 mm) in the 3‐D CT images (P < 0.05).

Table 1.

Comparison of the parameters between different image modalities (mean ± SD)

Parameters	Plain radiographs	3‐D CT scans	t‐value	P‐value
Initial NSA (°)	133.6 ± 3.8	131.4 ± 3.4	10.463	<0.05
Final NSA (°)	130.0 ± 1.9	128.8 ± 1.7	6.814	<0.05
Change NSA (°)	3.6 ± 2.8	2.5 ± 2.5	8.264	<0.05
Initial HHH (mm)	17.9 ± 0.9	16.8 ± 1.2	9.432	<0.05
Final HHH (mm)	15.8 ± 1.5	14.5 ± 1.1	7.219	<0.05
Change HHH (mm)	2.0 ± 1.6	2.3 ± 1.3	2.247	<0.05

HHH, humeral head height; NSA, neck shaft angle; SD, standard deviation.

Difference in Assessment of Reduction Loss

In evaluation of the reduction loss, the percentage was 16.9% (11/65) for the plain radiographs and 7.7% (5/65) for the 3‐D CT scans (P < 0.05). For the five patients with reduction loss, which was observed using the two imaging modalities, the mean change of the NSA and of the HHH was 9.18° ± 1.79°, 5.60 ± 0.56 mm in the plain radiographs and 8.20° ± 2.49°, 5.30 ± 0.48 mm in the 3‐D CT images, respectively. The mean Constant–Murley score27 of these 5 patients was 61.0 ± 1.6 (i.e. satisfactory). The mean Constant score of the 6 patients with reduction loss, which was observed in plain radiographs but not 3‐D CT images, was 82.7 ± 1.0 (i.e. indicating good; Fig. 2).

Figure 2.

The difference between plain radiographs and 3‐D CT images in the assessment of the postoperative reduction loss of a 65‐year‐old woman with a two‐part proximal humeral fracture treated with proximal humerus internal locking system (PHILOS) plates. Based on plain radiography, the change of neck shaft angle (NSA) and the change in humeral head height (HHH) were 10.1° and 5.1 mm, respectively, which was the difference between the postoperative initial (A and E) and 6‐month (B and F) measurement, showing reduction loss. Based on 3‐D CT, the change of NSA and the change of HHH were 7.6° and 4.7 mm, respectively, which was the difference between postoperative initial (C and G) and 6‐month (D and H) measurement, showing no reduction loss. The shoulder function was good (Constant–Murley score: 84 points). α is the head inclination angle.

Complications

Three major complications were noted: screw cutout, varus collapse, and subacromial impingement. Screw cutout was observed in 1 patient on the 3‐month follow‐up radiograph. This patient was advised to undergo refixation, but she refused any treatment. Varus collapse was seen in 2 patients, 12 and 11 months after surgery. These 2 patients underwent arthroplasty. Subacromial impingement was found in 1 patient on the 2‐month follow‐up radiograph, and he underwent refixation. None of the patients had other complications, such as humeral head avascular necrosis (AVN), infection, or adhesive capsulitis.

Inter‐observer and Intra‐observer Reliability

All values of intra‐observer and inter‐observer reliabilities of the variables are shown in Table 2. All ICC values of the 3‐D CT measurements (ICC range, 0.834–0.967) exceeded 0.8, indicating excellent agreement; these values were higher than those for the plain radiological measurements (ICC range, 0.598–0.915).

Table 2.

Intra‐observer and inter‐observer reliability of two image modalities for the measurements

Image modalities	Intra‐observer		Inter‐observer
	ICC	95% CI	ICC	95% CI
Plain radiographs
Initial NSA	0.795	0.643–0.879	0.754	0.616–0.843
Initial HHH	0.808	0.680–0.878	0.753	0.615–0.855
Final NSA	0.755	0.613–0.854	0.742	0.598–0.847
Final HHH	0.853	0.741–0.915	0.789	0.612–0.891
3‐D CT scans
Initial NSA	0.935	0.817–0.966	0.924	0.878–0.956
Initial HHH	0.912	0.834–0.954	0.910	0.852–0.945
Final NSA	0.923	0.845–0.957	0.919	0.862–0.954
Final HHH	0.934	0.878–0.967	0.913	0.843–0.957

CI, confidence interval; HHH, humeral head height; ICC, intraclass correlation coefficient; NSA, neck shaft angle.

Discussion

Reduction loss is a common complication in the treatment of proximal humeral fractures using locking plates11, 14. Postoperative NSA, HHH, and their changes are the prognostic factors for reduction loss11, 13. Therefore, it is crucial to find a reliable radiographic technique to evaluate the above prognostic factors. In the assessment of NSA and HHH, the examiners obtained excellent intra‐observer and inter‐observer reliability using 3‐D CT images (ICC range, 0.834–0.967). Significant differences were found between the two image modalities in the evaluation of the prognostic factors. Meanwhile, there were significantly different results for the percentage of reduction loss between plain radiography and 3‐D CT. Six patients with reduction loss, which was observed only in plain radiographs but not 3‐D CT images, had good shoulder function. These results of the present study showed that 3‐D CT was more reliable than plain radiography in the assessment of reduction loss. The reasons why plain radiographs were inaccurate for measuring NSA and HHH included projection errors, poor compliance, and image overlap. However, these factors, which occurred in the radiographs, did not affect the CT scans. The advantages of CT have been confirmed in the preoperative planning for proximal humeral fractures28. In the postoperative assessment of proximal humeral fractures, its values have not been evaluated.

The neck shaft angle is the important radiographic factor in the evaluation of postoperative outcomes for proximal humerus fractures. A previous study demonstrated that the average NSA of normal humerus was 134.7°29. Prior clinical investigation revealed that the mean postoperative NSA was 130° using plain radiographs14. Based on the same technique, Pawaskar et al. found the mean NSA to be 133.6° initially postoperatively, and 128.4° at final follow‐up30. Compared to normal humeri, there was, to some extent, a decrease for postoperative humeri. These findings were comparable to the present study that reported that the mean NSA was 133.6° initially postoperation and 130.0° at final follow‐up based on plain radiographs. However, the mean NSA initially postoperation was 131.4° and at final follow‐up was 128.8° using 3‐D CT14, 30. This difference may be due to the position of the arm with regard to the beam, which has been proven by Bai et al. who confirmed that NSA was smaller for an externally rotated glenohumeral joint, and larger for an internally rotated joint16.

It has been reported that initial postoperative NSA (<120°) is a risk factor for the functional outcome14. Zhang et al. showed that when the initial postoperative NSA was >127.9°, the incidence of loss of internal fixation would decrease17. Thus, postoperative NSA was an effective predictor for the functional outcome. However, further studies are needed to clarify these patients’ outcome where the initial postoperative NSA was from 120° to 127.9°. The change in NSA from initial postoperative to final follow‐up is also a prognostic factor. Previous studies showed that the change in NSA (≥10°) indicated poor functional outcome and reduction loss that would result in varus malunion11, 15, 16. Bai et al. demonstrated that in the reduction loss group, the mean change of NSA was 12.6° ± 3.1° and the mean Constant score was 64.1 ± 9.7, which had some differences to the present study for the change in NSA (mean change of NSA: 9.18° ± 1.79° in plain radiographs, 8.20° ± 2.49° in 3‐D CT images; mean Constant score: 61.0 ± 1.6)16. The differences could be explained by HHH ≥5 mm but NSA<10° change for some patients.

The humeral head height plays an important role in the assessment of the functional outcome and varus malunion of proximal humeral fractures, which is similar to the function of the neck shaft angle. For normal humeri, the “humeral head height” was defined as the vertical distance between the most superior point on the greater tuberosity and the most superior point on the humeral head; the mean humeral head height was 7.96 mm29. There was a positive correlation between “humeral head height” and NSA31. Special attention should be paid to the position of the PHILOS plate; the distance from the most superior point on the plate to the most superior point on the great tuberosity was 5–10 mm13. HHH was an important parameter, because too small a value could result in subacromial impingement, and too large a value could affect the placement of medial support screws. An early clinical study revealed that the mean change of HHH was 1.6 mm18, which concurs with the results of the present study. The change of HHH was the risk factor for reduction loss, and the change (≥5 mm) could lead to poor shoulder function11, 16, which concurred with the findings of the present study (mean Constant score of the patients with reduction loss: 61.0 ± 1.6). The present study showed that the mean change of HHH in the reduction loss group was 5.6 mm in the plain radiographs and 5.3 mm in the 3‐D CT images, which was similar to the previous study (mean change of HHH: 5.0 ± 2.5 mm)16. It has been reported that there was also a positive correlation between the change of HHH and the change of NSA16. This further proved that the change of NSA and HHH were both prognostic factors for poor shoulder function. Therefore, it is crucial to identify a reliable radiographic technique that can precisely evaluate the prognostic factors. This may determine the subsequent clinical management, which might involve further surgery or more frequent follow‐up examinations.

There were some limitations in this study. First, the present study only included two common parameters we believed important for evaluating postoperative functions. Other parameters could be also selected for the evaluation, and further studies are required. Second, the present study was not a multi‐center study. The results of the fracture reduction and internal fixation were related to the level of clinical skill of the present team.

The present study demonstrated that 3‐D CT images were more reliable than plain radiographs in the assessment of the prognostic factors of reduction loss of fractures of the proximal part of the humerus with the treatment of locking plates. Therefore, we believe that this reliable technique can serve as an effective guide for subsequent clinical management of patients.