Kyphoplasty was performed in cases wherein conservative treatment failed to alleviate the pain in 6 weeks after the appearance of the symptoms. The kyphoplasty technique was chosen according to the spine surgeon’s choice, without any specific preference for a particular procedure. Each procedure was performed after obtaining written informed consent from the patient. Patients were positioned in the prone position. Closed reduction before the procedure was attempted by using displaceable positioning pillows or flexible surgical tables. A bipedicular surgical approach was performed for BKP using the KYPHON (Medtronic plc, Dublin, Leinster, Ireland) or SYNFLATE (Depuy Synthes, West Chester, PA, USA) system. In contrast to BKP, a unipedicular approach was used for RFK with the StabiliT system (DFine, San Jose, CA, USA).

In addition to demographic data, pain intensity was measured using a Visual Analogue Scale (VAS) ranging from 0 (no pain) to 10 (worst pain) [12]. Preoperative and postoperative images were analyzed and gauged on the basis of lateral radiographs and/or sagittal planes of computed tomography. In accordance with the internal guidelines, lateral radiographs were acquired in a standing position and computed tomography was performed in the supine position routinely. The modified bisegmental Cobb angle and the vertebral angle between the intersecting lines of the top and bottom vertebral end plates were used as measures of the intervertebral kyphotic angle (Fig. 1) [13,14]. We further analyzed the anterior (Ha), middle (Hm), and posterior (Hp) vertebral body heights i.e., the distance (mm) between the top and bottom vertebral end plates perpendicular to the end plate [14]. The absolute and relative height restoration was calculated according to McKiernan’s methods A and B (Fig. 1) [15].

An individual magnification factor (MF) and follow-up date (FU) were calculated for each patient to account for the different magnifications of the X-rays resulting from varying focus-film (conventional) or source-image (digital) distances according to the following formula. Thus, all the measurements could be scaled comparable to the preoperative measurements (PO).

Statistical analysis was performed using IBM SPSS ver. 22.0 (IBM Corp., Armonk, NY, USA) and an analysis of variance was performed. Gaussian distribution was analyzed using the Shapiro–Wilk test. Statistical analysis included parametric tests (Student t-test) and non-parametric tests (Mann–Whitney U-test, Wilcoxon signed-rank test). Fisher’s exact test was used for analyzing contingency tables. Furthermore, logistic regression was performed. In case of binomial logistic regression, dependent variables had to be dichotomized, if necessary. For the evaluation of a linear connection between fracture age and height restoration as well as pain relief and height restoration, bivariate correlation analysis (Spearman’s rank correlation coefficient) was used. The significance level of p-value was set to 0.05. This research was published in accordance with STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines for observational studies.

Overall, 156 patients with a mean age of 73±11 years were included. For further information, please refer to Table 1. Although all the 156 patients kept the first follow-up appointment, clinical outcome data of only 155 patients was available. At 6 weeks after the operation, clinical data of only 56 patients were available, and at 1 year after the operation, that of only 10 patients were available. The radiological outcome data of 156 patients with 252 fractures was obtained after the operation, while after 6 weeks, the radiological outcome data of only 86 patients with 138 fractures was obtained, and at 1 year after operation, that of only 20 patients with 32 fractures was obtained because many were lost to follow-up. The mean fracture age at the time of the surgery was 54±83 days; 131 patients (84%) had a trauma prior to the fracture, 89 (57%) had osteoporosis, and 39 (25%) had an evidence of tumor in the fractured vertebral bodies. Demographic data did not differ between BKP and RFK groups (Table 1).

The mean preoperative pain intensity using the VAS was 7.6±1.6 and 2.4±1.8 after operation (p<0.001). Postoperative pain relief (ΔVAS) averaged 5.1±1.8 points. At 6 weeks after operation, the pain intensity averaged 3.1±2.5 (ΔVAS, 0.2±2.5; p=0.6), and at 1 year after the operation, the pain intensity averaged 2.9±2.8 (ΔVAS, 0.3±1.6; p=0.6). Table 1 gives an overview of the pain intensity at different times. In summary, there was no long-term change in pain intensity after the immediate pain relief that was felt after the operation (p=0.9). There were no significant differences in the surgical effects between BKP and RFK groups after the operation (p=0.2) at 6 weeks (p=0.5) or at 1 year after the operation (p=0.4). Moreover, the pain relief was not dependent on the fracture age (<6 versus ≥6 weeks) or the number of treated vertebral bodies (1 versus 2–5) at any time. For an overview of the pain relief, please refer to Table 2. With regard to the patient’s sex, men experienced a marked pain relief at 6 weeks after the operation (p=0.008). To probe the cause of this observation, we compared the age and fracture age between men and women. In this cohort, women were older than men (mean age, 74.3±10.7 years versus 68.8±11.9 years, respectively; p=0.006). The mean fracture age among men and women were 75.6±109 days and 42.9±65 days, respectively (p=0.1). Due to the medium-term different clinical outcomes dependent on sex, we dichotomized the patient population according to the median age (73.5 years). Patients in the younger group benefited more in the longterm (younger versus older patient group: 6 weeks ΔVAS, −0.4±1.9 versus 0.6±2.9; p=0.02; 1 year ΔVAS, −0.2±0.4 versus 1.0±2.6; p=0.01), while the differences immediately after the operation were not significant (ΔVAS, −5.4±1.7 versus −4.9±1.8; p=0.1). As the mean age of the treated men was markedly low, we analyzed the gender distribution within the young and old groups. The proportion of men was significantly higher in the young group (35/77 [45%] versus 16/78 [21%], p=0.002). To assess the influence of sex, age, and fracture age on the clinical outcome, we performed binomial logistic regression with postoperative pain relief as the outcome parameter (<3 versus ≥3). Nagelkerke’s pseudo R² value was 0.06 after the operation and 0.14 at 6 weeks after the operation. Only the male sex could be ascertained as an independent prognostic factor for a favorable clinical outcome postoperatively (VAS <3) with an odds ratio of 2.6 (95% confidence interval [CI], 1.2–5.4; p=0.01). At the 6-week follow-up, the odds ratio was 3 (95% CI, 0.8–10.6; p=0.09). Age as well as fracture age had no influence at any time.

Table 1 gives an overview of the postoperative restoration of vertebral body height. Height restoration was relatively small but significant with a change of vertebral angle of −1.3°±3.3° (p<0.001) and Ha, Hm, and Hp vertebral body heights (Ha, 1.5±2.9 mm; p<0.001; Hm, 2.1±2.9 mm; p<0.001; Hp, 0.9±2.1 mm; p<0.001). Change in Cobb angle (−0.3°±4.4°) was not significant (p=0.1). BKP and RFK showed comparable results. At the 6-week follow-up, there was a significant loss of reduction in the bisegmental Cobb angle (1.3°±4.6°, p=0.002) and Ha vertebral body height (Ha, −0.9±3.0 mm; p<0.001). The change in the vertebral angle, and Hm as well as Hp body height were not significantly different between BKP and RFK. At the 1-year follow-up, we observed a further slight loss of reduction primarily at the anterior vertebral body edge (Ha, −0.9±2.2 mm; p=0.04; Hm, −0.8±1.6 mm; p=0.01). There was no significant change in the vertebral angle, Cobb angle, and Hp vertebral body height. In summary, both kyphoplasty methods achieved comparable postoperative height restoration. However, height restoration could not be maintained over the follow-up period, irrespective of the applied kyphoplasty method.

There was no correlation between fracture age and height restoration (Spearman rho, −0.08 to 0.04; p>0.05), pain relief and height restoration (Spearman rho, −0.02 to 0.03; p>0.05), and pain relief and fracture emergence (Spearman rho, 0.04 to 0.15; p>0.05).

Average postoperative pain relief was 5.1±1.8, and this is in accordance with the published literature. Every single study that measured pain relief using VAS reports a significant postoperative pain reduction. The pain reduction averaged around 5 in recent studies on kyphoplasty. Our results confirm this score of pain relief and support the assumption that kyphoplasty is an effective treatment option for vertebral compression fractures. However, the mean pain relief varies considerably in different studies from 1.2 observed by Reinhold et al. [16-18] to 7.6 obtained by De Negri et al. [19] at 1 year after the operation. Studies that compare BKP and RFK are scarce. Bornemann et al. [9] performed a matched pair analysis with 96 patients in each group and reported a significantly increased pain reduction in the RFK group at 12- and 24-month follow-up (VAS 1.2 versus 2.1). There was no difference in the extent of pain relief immediately after the surgery between the two methods [9]. Pflugmacher et al. [8] reported similar results with a pain reduction of 7.3 and 5.9 using BKP and RFK, respectively, at the 6-month follow-up (p<0.001), while there was no difference immediately after kyphoplasty (5.9 versus 5.5, p=0.2). Petersen et al. [10] reported a comparable pain reduction with BKP and RFK (−4.6 versus −4.4) with lower pain at rest after 1 year in the RFK group. These results suggest that RFK results in long-term pain relief. Our results demonstrate favorable outcomes with both BKP and RFK. However, our results do not allow a reliable assessment of the long-term outcome because of the small sample size in the RFK group as well as because many patients were lost to follow-up.

Height restoration at the anterior vertebral body edge and the center of the vertebral body differs broadly among studies. Our results (1.5 and 2.1 mm) are at the bottom end of all studies and are comparable to the findings of Santiago et al. [20], Röllinghoff et al. [21], and Hiwatashi et al. [22]. The most favorable height restoration was reported by Liu et al. [23] with a mean absolute gain in height of 9.1 mm in 50 patients treated with BKP. In a study of Liu et al. [23], preoperative Ha was remarkably low (11.3 mm) and only recent fractures (mean, 17 days) had been included, which might explain for the extraordinary gain in height.

The most favorable angle-based results were reported by Yan et al. [24] with a mean change of kyphotic angle of 11.7°. Our results are comparable to that of Kasperk et al. [25] who reported a mean change of kyphotic angle of 1.1° and that of Röllinghoff et al. [26] who reported a mean change of 1.0°. However, the comparison of angle-based height restoration is challenging because of the imprecise definition of kyphotic angle in different studies. Specific comparisons between BKP and RFK have been performed by Pflugmacher et al. [8], Bornemann et al. [9], and Petersen et al. [10]. Each study displayed slightly better but insignificant numbers in terms of reposition in the RFK group [8-10].

Our study has the following limitations. A retrospective study design can facilitate a selection bias. Furthermore, a limited sample size and a serious loss to follow-up additionally hamper statistical validity. There is no reliable information on the extent to which closed reduction contributed to the postoperative reduction. Therefore, there might be a confounding effect of closed reduction independent of the applied kyphoplasty method. We are convinced that an individual MF may enable one to compare medical images with higher validity. Nevertheless, scaling images with a standardized radiopaque body as a benchmark should be the method of choice.