Gastric cancer (GC) ranks among the most prevalent and fatal malignancies globally, with around 1.09 million new cases and 768,000 deaths reported in 2020 [1]. The main treatment option for localized GC is surgery. However, the risk of recurrence and metastasis remains high, particularly in advanced stages [2,3]. Various strategies, including perioperative chemotherapy and adjuvant chemoradiotherapy (CRT), aim to improve long-term survival [2]. While the benefit of radiotherapy (RT) as an adjuvant treatment is debatable, studies show no significant overall survival improvement with RT after quality-assured gastrectomy [4-6]. Adjuvant CRT could improve overall survival for patients, especially those who have undergone D1 lymphadenectomy or R1 resection [7-9].

GC patients undergoing gastrectomy face elevated osteoporosis and fracture risks compared to the general population, with incidences of 36% and 42.1 cases per 1,000 person-years, respectively [10,11]. Cancer treatments like endocrine therapies and chemotherapy are known to reduce bone mineral density (BMD) and increase fracture risks, but the association with RT remains unclear [12,13]. RT can induce bone loss and fragility by directly damaging the bone cells and matrix [14,15], as well as indirectly affecting the hormonal and inflammatory pathways that regulate bone metabolism [16,17]. RT could have negative effects on the bone health of GC survivors, such as decreased BMD, osteoporosis, and vertebral compression fractures (VCFs) [18,19].

VCF is a collapse of the vertebral body that can cause pain, disability, deformity, and increased mortality. VCFs can occur spontaneously or as a result of minor trauma in patients with weakened bones [20]. Several factors have been identified as potential predictors of VCFs in cancer survivors, such as age, sex, body mass index, smoking, alcohol intake, history of osteoporotic fracture, osteoporosis, rheumatoid arthritis, diabetes mellitus, hyperthyroidism, systemic glucocorticoid usage, low BMD, and low Hounsfield units (HUs) [14,21-26]. Treatment-related factors like higher RT doses, technique, surgery type, and chemotherapy usage may also influence VCF risk [14,22,26,27]. VCFs have been reported in a variety of cancers following RT, including cervical, uterine, prostate, rectal, lung, pancreatic, esophageal, and GCs [18,24-26,28-36].

HUs, quantifying material radiodensity on computed tomography (CT) scans, are associated with BMD and predict osteoporosis and VCFs in diverse populations. [19,37-40]. Yet, the utility of the HU in predicting VCFs after adjuvant irradiation for GC has not been thoroughly investigated.

This study aims to investigate the correlation between HU and VCF occurrence in GC patients receiving adjuvant RT or CRT, along with exploring other potential risk factors for VCFs. We hypothesized that lower HU values would correlate with higher VCF rates, and the HU could be a simple and convenient way to assess VCF risk in GC survivors.

The median age at diagnosis was 58 years (range, 20 to 83). The detailed patient, tumor, and treatment characteristics are presented in Table 1. The most frequently applied neoadjuvant chemotherapy regimen was docetaxel, cisplatin, and 5-fluorouracil in nine patients (45.0%), followed by 5-fluorouracil, folinic acid, oxaliplatin, and docetaxel in eight patients (40.0%). The most frequently used concomitant chemotherapy regimen was capecitabine in 148 patients (58.5%). The most frequently applied adjuvant chemotherapy regimen was capecitabine and oxaliplatin in 79 patients (31.2%), followed by 5-fluorouracil and folinic acid in 63 patients (24.9%).

The median follow-up time was 35.7 months (range, 2.8 to 136.5). VCFs developed in 23 out of 271 patients (8.5%). The median development time for VCFs was 30.6 months (range, 3.4 to 117.3). Six patients (26.1%) developed fractures within the first year, and 11 patients (47.8%) within 2 years. Five patients (21.7%) developed fractures after 5 years. Nine patients (39.1%) had VCFs at multiple vertebral levels. Of these, five had fractures at two levels, three had fractures at three levels, and one had fractures at four levels. VCF was observed in 37 vertebrae. Fourteen of them (37.9%) were located at T12 level, nine (24.3%) at L1 level, seven (18.9%) at T11 level, four (10.8%) at L2 level, and three (8.1%) at T10 level. All fractured bones were within the irradiated volumes.

The effects of clinical features on the development of VCFs are shown in Table 2. Older age (p = 0.007), female sex (p = 0.021), and non-smoking status (p = 0.023) are associated with a higher incidence of VCFs. The effects of vertebral doses and HUs on the development of VCFs are shown in Table 3. Patients who developed VCF had significantly lower median HUs for T10, T11, T12, L1, and L2 vertebrae (212.9, 197.2, 182.4, 187.6, and 188.3) compared to those who did not (239.3, 224.5, 217.6, 208.8, and 214.7), respectively (p = 0.005, p = 0.003, p = 0.002, p = 0.009, and p = 0.003). The development of VCFs was not significantly affected by the history of alcohol intake, surgery type, RT technique, RT total dose, or the doses of radiation received by the vertebrae (V30, V35, V40, and D_mean).

Univariate and multivariate analyses were performed for factors found to be associated with VCF development (Table 4). Older age, female sex, non-smoking status, and lower median HUs of T10-L2 vertebrae were significantly associated with a higher probability of developing VCFs in the univariate analysis. Only lower median HUs of T12 vertebrae (odds ratio, 0.965; 95% CI, 0.942 to 0.989; p = 0.004) remained significantly associated with VCF development in the multivariate analysis.

In the analysis of whether age (<58 vs. ≥58 years) and sex (male vs. female) affect T12 HU values, it was found that neither sex (p = 0.499) nor age (p = 0.151) had a significant impact on the T12 HU values. The impact of T12 HU values on VCF incidence stratified by age and sex is shown in Table 5. In all subgroups, patients with VCF had significantly lower T12 HU values compared to those without VCF.

ROC analysis was performed to find the cut-off value for T12 HUs that best predicted VCF development (Fig. 2). The ideal cut-off value for T12 HUs forecasting VCF development was found to be 205.1. Area under the ROC curve for T12 HUs was 0.765 (95% CI, 0.645 to 0.886; p = 0.002). The optimal sensitivity and specificity for the T12 HUs cut-off level were 85.7% and 65%, respectively. For a threshold value of 221 HUs, the sensitivity was found to be 93%, and for a threshold value of 171.5 HUs, the specificity was found to be 90%.

To our knowledge, this is the most extensive research on VCFs following adjuvant RT or CRT for GC, comprising the data of 271 patients. In this study, T12 vertebrae HUs were associated with the risk of VCF development in GC patients who underwent RT. The cut-off value was determined to be 205.1, with 85.7% sensitivity and 65% specificity.

Numerous studies have explored VCF prevalence and associated risk factors, revealing incidence rates ranging from 1.7% to 89% in patients undergoing treatment for various cancers such as gynecologic, gastrointestinal, lung, pancreatic, or prostate cancer [14,34-36]. Various methodological differences, such as study designs (retrospective or prospective), varying follow-up scan frequencies, different imaging techniques, RT doses, and the inclusion of asymptomatic VCFs, collectively account for the observed wide range. Our study reported a VCF rate of 8.5%, which is consistent with the literature [18,28,33-36]. We observed that the vertebral fracture rates in studies examining VCF in patients receiving RT due to cervical cancer in Japan were higher than the fracture rate in our study [24,25,30,31]. We think that this could possibly be explained by the population being predominantly female, known to have higher fracture risks, and the higher prevalence of vertebral osteoporosis in Japanese society compared to the United States and Europe [14,44].

In our study, the median time for the development of VCFs was 30.6 months, longer than previously reported in the literature, which ranged from 6 to 20 months [14]. This discrepancy may be attributed to the extended follow-up period in our study. Among patients who developed VCF, 21.7% experienced fractures after 5 years. The longest VCF development time observed was 117.3 months, which, to our knowledge, is the longest duration reported for VCF development in the literature. Most VCFs (62.2%) occurred in the T12 and L1 vertebrae, which are junctional and weight-bearing, mirroring findings in cervical cancer patients in whom fractures frequently appear in the weight-bearing sacrum and sacroiliac joints [29]. These findings indicate that even though RT affects all bones in the treatment field, the weight-bearing vertebrae are the most susceptible to fracture development.

In evaluating factors influencing VCF development, older age, female sex, non-smoking status, and low vertebral HU values were associated with increased VCF incidence in univariate analysis. Subsequent analysis revealed a significant association between smoking status and sex (χ² = 12.72; p < 0.001). We concluded that the relationship between non-smoking status and VCF formation was mainly due to sex, as most males (61.0%) were smokers and most females (61.7%) were non-smokers, which indicates smoking status as a confounding factor. Similar to our findings, numerous studies on VCF risk factors in cancer patients have also linked advanced age and female sex with higher VCF incidence [26,33,35].

The pathophysiology of radiation-induced bone damage remains incompletely understood. Irradiation lowers the number of osteoblasts, stops their cell cycle, and makes them more prone to cell death, meaning that less bone formation is a key factor in bone damage caused by RT. Radiation also seems to harm the bone matrix, increase fat in the marrow, and impair the blood supply to bone, all of which likely affect the direct impact of RT on bone health and strength [14]. Our study found no association between total RT dose, vertebral doses (V30, V35, V40, and D_mean), and VCF development, despite conflicting literature data. Studies show that every 5-Gy increase in the mean vertebral dose raises the risk of VCF development by factors of 1.19 and 1.22, according to Fujii et al. [26] and Wu et al. [36], respectively. In both studies, the median total RT dose was 60 Gy. This exceeded the median total RT dose in our study, which was 45 Gy. In addition, Uezono et al. [24] examined the risk factors for pelvic fractures in cervical cancers that received RT. They also found no significant association between fractures and vertebral doses, consistent with our study. The median total RT dose was 50.4 Gy, which closely resembled that of our study. Considering all these data, it was thought that the RT dose threshold for vertebral bones could be so high that we could not observe the effect of the RT dose on VCF development.

Osteoporosis, which is indicated by low BMD, is a significant risk factor and an indirect predictor of VCF development [14]. While dual-energy X-ray absorptiometry (DEXA) is considered the gold standard for assessing BMD [45], it is not routinely used for cancer survivors. This, coupled with the focus of BMD measurements on areas like the lumbar spine and femoral neck, makes it difficult to assess bone loss in radiation-affected areas. Hence, considering bone density measurements using CT before RT could be valuable in predicting the development of fractures. Numerous studies have shown that there is a strong relationship between BMD assessed by DEXA and HUs derived from CT scans [19,37-40]. One of the most important findings of our study was that pre-RT HU values of vertebrae were the best parameter to show the probability of VCF development, as well as the findings of previous studies [24,25]. While all vertebral HU values were associated with VCF development in univariate analysis, only the T12 HU value remained significant in multivariate analysis. Uezono et al. [24] and Ishikawa et al. [25] conducted studies investigating the risk of pelvic fractures in patients with cervical cancer who received definitive RT. In both studies, it was observed that only pre-RT CT HU values were associated with the development of pelvic fractures, consistent with our study.

We identified 205.1 as the optimal cut-off point for the T12 vertebrae HU value. A threshold of 221 HUs showed 93% sensitivity, and 171.5 HUs had 90% specificity. Wu et al. [36] found a higher risk of thoracic vertebral fracture in esophageal squamous cell carcinoma patients with initial CT scan HU values below 128. However, their study did not include the cortical edges of vertebrae in the region of interest for HU determination, possibly accounting for the discrepancy.

Our study aimed to simplify VCF risk assessment for radiation oncologists. The HU values of the vertebrae indicated the likelihood of VCF occurrence. Therefore, the VCF risk could be estimated by using treatment planning systems. We can also use the HUs from CT scans to estimate BMD, save money, and reduce radiation exposure from DEXA tests. Moreover, the sagittal views of the vertebrae from CT scans can be used for detecting hidden and asymptomatic bone fractures. By incorporating these measurements into regular practice, osteopenia, osteoporosis, and fracture risk could be identified.

Our study has several strengths, such as a relatively large sample size, a long follow-up period, a homogeneous population of GC patients, a standardized adjuvant RT or CRT protocol, and objective measurement of HUs and VCFs. However, our study also has some limitations, mostly due to its retrospective nature. The patients’ comorbidity status, body mass index, basal calcium level, and vitamin D level, as well as whether they received replacement therapy during follow-up, were unknown. The patients’ baseline BMD was not assessed using the DEXA method, which is the most reliable technique. The clinical outcomes of VCFs, such as pain, disability, quality of life, and mortality, that are important aspects to be considered in the management of cancer patients, were not evaluated. As all patients in our study population received adjuvant RT, we were unable to compare the development of VCF and associated factors between patients who received RT and those who did not. Furthermore, the effect of RT on the HU value was not observed by measuring HUs in the sequential CT scans of the patients during the follow-up. Finally, since there are several techniques for HU measurement, cut-off values for HUs should be interpreted with attention to the measurement technique used in each study.

In conclusion, our study suggests that GC patients who receive adjuvant RT or CRT are at increased risk of VCF. Additionally, low pre-RT or CRT vertebral HU values are simple and reliable indicators of VCF risk. These results imply that HU measurement from CT scans can be used as a screening tool to identify high-risk patients who may benefit from preventive interventions, such as replacement therapy with calcium and vitamin D, vertebral augmentation, or prophylactic vertebroplasty. Therefore, further studies with prospective, long-term, and comprehensive assessments are needed to validate and extend our findings and to explore the possible mechanisms and preventive strategies for VCFs in GC patients who receive RT.