This study showed that 3D-VR was accurate in measuring shortening of midshaft clavicle fracture, with mean difference of 0.736 mm (95CI = − 2.51;3.98) compared to manual measurement. When compared to 2D radiographic methods, a partial correlation was observed. The method proposed by Silva et al. and the HUG method were the only measurements correlating significantly with the CT measurements – the HUG method being the less sensitive to false-positive surgical indication. There was no evidence of impact of fracture pattern on shortening.
Historically, midshaft clavicle fractures have been treated conservatively with low nonunion rate, but without including displacement factors [1, 19]. More recently, some studies have shown less satisfactory results with nonunion rates up to 15%, depending on the displacement, with an impact on the function of the shoulder [8, 20]. This has led to highlight several risk factors for poor results of conservative treatment [21]. Among those, shortening has been shown to have an impact on the union rate, function, and personal satisfaction [4, 22,23,24]. McKee et al. showed an inverse relationship between shortening and endurance in abduction for a threshold of 20 mm of shortening [23]. This trend was confirmed by Lazarides et al. which showed poor functional results with a threshold of 14 mm for women and 18 mm for men [22]. It seems that the clavicle shortening is reflected throughout the shoulder with impaired scapula-thoracic and glenoid alignment, which negatively affects mobility and muscle tension in the peri-scapular musculature [25,26,27,28,29].
Thus, the knowledge of the effect of shortening of clavicle fractures have led to more indication to osteosynthesis of these fractures. Better results were obtained for functional scores, accelerated return to work and overall satisfaction when a surgery was proposed for displaced fractures [22,23,24]. However, this led sometimes to the over-indication of surgery along with possible complications, such as infection and necessity of hardware removal.
To avoid over-indication, the goal of this study was to develop accurate measurements of clavicle fractures shortening in daily practice. CT is widely recognized in the literature as the gold-standard in shortening measurement [10, 16]. This study confirms the use of CT with virtual fragments reposition as the reference modality, with a non-significant minimal difference of 0.736 mm (95CI = − 2.51;3.98) compared to manual clavicle length measurements. However, in clinical practice, the surgeon is frequently limited to decision based on 2D imaging. Omid et al. assessed the measurement of standard radiographic shortening compared to CT [10]. Despite good reproducibility of radiological measurements, they were not significantly correlated with those on CT. Nevertheless, in this study, measurements were made on chest X-rays by measuring the length difference with the contralateral side. This methodology assuming symmetry between the two sides may be biased. Cadaveric and anthropological studies have shown side to side differences up to 15 to 20 mm [12, 13]. More recently, Cunningham et al. performed a radiological study comparing the length of the two clavicles on CT [30]. They demonstrated 28% asymmetry of more than 5 mm, of which 7% of more than 10 mm. This was confirmed by Hoogervorst et al., showing 30% of asymmetry over 5 mm when measured on CT [31].
In the present study, four measurement methods were evaluated based on single side imaging. The Silva et al. and the HUG methods were significantly correlated to 3D measuring with mean difference of − 1.02 mm and − 2.03 mm, respectively, therefore these may be used clinically. This is confirmed by the Bland-Altman analysis as shown in Fig. 3. These two methods are those whose differences with CT are closest to zero with minimal dispersion. Jones et al. measured shortening on a single incidence [32]. Their analysis focused solely on the reproducibility of the measurements (without specifying the technique). Their results showed a low inter-observer correlation. On the opposite, our study showed a very good reproducibility with ICC from 0.84 to 0.98. However, the methods of Jeray and Smekal et al. showed a limited correlation with the CT measurements, and we do not recommend their use for clinical decision.
The technique based on the study of Silva et al. showed measurements performed on anteroposterior incidences of fracture of the middle third of clavicle of pediatric patients [15]. The analysis focused solely on the reproducibility of measures, without correlation with CT. The inter-observer reliability was not good enough to allow its clinical use (ICC 0.69 to 0.74). In our study, we applied this method to adult clavicles with good reliability (ICC 0.97). When compared to 3D CT reconstructions, the results were also reliable. The HUG method was the measurement showing the best correlation with 3D-VR among the four radiographic methods evaluated.
In addition, the method of Silva et al. and the HUG method were the ones that minimized the most the diagnostic error. A difference exceeding 20 mm of shortening in comparison with 3D-VR was found in 5–10% (two observers) with the HUG method and 15% with the Silva et al. method. Use of 3D-VR technique could hence avoid unnecessary surgical treatment and related complications for the patients.
Strength of our study lies in the use of 3D CT reconstruction. Indeed, most of the studies comparing standard radiographic measurement with CT remained in the 2D plane [10, 16]. However, 3D reconstruction provides the advantage to standardize the reference points for the measurement of length, and thus to minimize the error rate. Even if the collimation used on a total body CT for a polytrauma patient is different from that of a clavicle, because of the wide field of view studied (soft tissue and bony structures), the infra-millimetric difference between cut of 1.25 mm and 0.3 mm probably does not affect the measurement significantly, especially on 3D-VR reconstructions. To our knowledge, this study is the first to evaluate shortening with 3D reconstructions and virtual repositioning of fragments, simulating surgical management. Moreover, all studies comparing 2D vs. 3D measurements were done on consolidated or non-fractured clavicle. Our study is the first to assess fractured clavicle.
Shortening has usually been measured by standard radiographs. Studies have shown that image distortion due to radiological image incidence modifies the amount of measured shortening, leading to changes from operative to nonoperative indication in 33.9% of the cases [33, 34]. With our study confirming the accuracy of the use of CT and 3D reconstruction with virtual reposition, this technique could be a valuable alternative, opening the question on the use of low-dose CT for clavicle fracture.