The purpose of this study was to generate real-world data, from US sources, regarding incidence and risk factors for subsequent VCF after vertebral augmentation of an incident fracture. We sought to understand how often subsequent treatment occurs and the factors that increase or reduce the risk that they will require subsequent treatment, with the goal of providing data that could help inform patient selection and patient choice.

Overall, our study found a subsequent treatment risk of 4.8% within the first 30 days of the index procedure, with the rate of subsequent treatment falling over time (9.4% at 90 days, 13.2% at 365 days, and 15.5% at 730 days after index). The greater risk in the first 30 days may be related to a number of factors including lack of advanced imaging, such as with MRI for patient selection to identify initially occult fractures, or time delay between initial diagnosis with advanced imaging to treatment and additional fracture prior to initial treatment. No advanced imaging data was available in our administrative dataset to explore this further.

Our subsequent treatment rate with BKP or VP observed within the first 2 years following a primary vertebral augmentation of 15.5% is consistent with the findings of the 2022 meta-analysis by Dai and colleagues [6], and roughly within the mid-range of numerous other studies published within the last 14 years [6, 24,25,26,27,28,29,30,31,32,33]. Interestingly, this rate is also similar to the results of a study published in 2001 by Lindsay and colleagues [34], who analyzed the natural history of VCF among 2725 postmenopausal women who had been randomized to the placebo groups in 4 large, 3-year osteoporosis treatment trials conducted at 373 centers in North America, Europe, Australia, and New Zealand from November 1993 to April 1998. Due to the design of the trials, baseline VCF status was known for all participants; the main outcome measure was radiographically identified vertebral fracture during the year following an incident vertebral fracture. This study found that the cumulative incidence of any VCF in the first year was 6.6%. Patients who entered the study with one or more prevalent VCFs at baseline were at approximately fivefold risk of a new VCF within the first year of the study compared to patients with no prevalent VCFs at baseline. Among the 381 patients who did develop a new VCF during the study period, the incidence of another vertebral fracture within the next year was 19.2% (95% CI 13.6–24.8%). Thus, the incidence of subsequent VCF after vertebral augmentation found in our study was similar to the historical incidence of subsequent VCF in patients managed conservatively and without surgical intervention.

In alignment with these findings are the results of a 2021 meta-analysis by Halvachizadeh and colleagues, which analyzed adjacent-level fractures and other outcomes from 16 randomized controlled trials evaluating vertebroplasty and kyphoplasty against each other or against non-operative management [21]. The analysis found no significant differences in the incidence of adjacent-level fractures between patients managed with vertebral augmentation (22.7%) or non-operative management (22.6%). Furthermore, the risk of adjacent-level fractures in the non-operative management group was comparable to that in the vertebroplasty groups (log OR = − 0.16; 95% CI − 0.83 to 0.50; heterogeneity, I2 = 72.5%).

When we analyzed factors that either reduced or increased the risk of subsequent VCF after vertebral augmentation, our findings were in line with known risk factors. For example, male sex, races other than White (and especially Black or Asian race), younger age, and overweight body habitus were protective in this study population.

Characteristics associated with significantly higher hazard of subsequent treatment included vertebral augmentation at more than one fracture level, comorbid arthritis, use of steroid medications, tobacco use, and multiple myeloma, each of which has precedent in the literature as a risk factor for subsequent VCF [2, 3, 5,6,7, 25, 26, 35,36,37]. Overall, the risk of subsequent treatment after an initial vertebral augmentation increased as overall patient comorbidity increased as demonstrated with increasing Elixhauser Comorbidity Index score. A somewhat surprising finding was the low prevalence of any osteoporosis medication prescription fill, both in baseline and in follow-up. Exploratory analysis revealed that only 6.6% of patients in follow-up were newly prescribed osteoporosis medication(s) after their initial VCF treatment. This suggests potentially low referral rates to endocrinology for continued management for osteoporosis, or patient adherence issues in filling and starting osteoporosis medication treatment.

Thus overall, our study confirms that patient comorbidities and initial fractures at baseline treatment are associated with the risk of subsequent fracture and need for subsequent vertebral augmentation. Patients with fewer comorbidities, of younger age, without cancer, and with fewer initially fractured and treated levels are significantly less likely to experience subsequent fracture after treatment. Patients with additional risk factors may also be suitable candidates, but the decision to treat VCF with vertebral augmentation should be balanced against that particular patient’s risk profile, the likelihood of benefit, and patients informed about the risk of subsequent fracture. Clinicians seeking additional guidance on how to select patients for vertebral augmentation are referred to a recent multidisciplinary consensus paper by Hirsch et al. [5].

A significant limitation of the study design was the inability to evaluate the occurrence of new VCF during the follow-up period directly, using VCF diagnosis codes. ICD-10-CM coding has a single diagnosis of VCF, not differentiating between new or subsequent fracture. Thus, it was necessary to use a proxy measure for refracture by evaluating visits in follow-up with additional/subsequent VCF procedure codes. Therefore, the rates of subsequent treatment reported in this study do not represent the overall repeat fracture rate, as some patients with subsequent fracture may have declined another surgery.

Furthermore, the ICD-10 diagnosis code for VCF is broad and non-specific and does not provide information with which to distinguish the specific spine level at which the initial and subsequent VCFs were treated (e.g., same or adjacent). Therefore, it was not possible to evaluate the influence of initial treatment level(s) on risk of subsequent treatment. In the future, prospective research could be designed that would allow differentiation between vertebral levels to evaluate risk of subsequent adjacent-level fractures and their relation to risk factors identified in our present study. Finally, as a retrospective analysis of commercial claims, the data are subject to coding errors and may not reflect the entire US VCF population.