Prostate cancer, a leading malignancy among men, is frequently associated with an increased risk of osteoporosis and subsequent osteoporotic fractures (1,2). These fractures contribute significantly to morbidity and mortality, particularly in elderly patients (3). Androgen deprivation therapy (ADT), a cornerstone treatment for prostate cancer, is a well-established risk factor for bone density loss, exacerbating the risk of osteoporosis. However, despite these risks, routine osteoporosis screening is not consistently implemented in this patient population, leaving many at risk for preventable skeletal events (4,5). Current osteoporosis screening methods, such as dual-energy X-ray absorptiometry (DEXA), are the standard for assessing bone mineral density. However, the effectiveness of DEXA is often compromised in patients with prostate cancer, especially those with osteoblastic metastases (6), which can lead to the overestimation of bone density. Additionally, DEXA scans are underutilized in clinical practice, particularly for patients with non-metastatic disease on ADT, due to logistical challenges and a lack of standardized guidelines. This underutilization highlights a critical gap in the management of bone health in patients with prostate cancer, necessitating the exploration of alternative diagnostic tools (7-9).

This study aims to address this gap by evaluating the utility of incidental cross-sectional imaging, such as computed tomography (CT) and positron-emission tomography (PET)/CT), in the diagnosis of osteoporosis among patients with prostate cancer. These imaging modalities, often performed for other clinical reasons, might offer a valuable, opportunistic screening tool for osteoporosis without the need for additional procedures. By analyzing the attenuation of the L5 vertebral body across multiple imaging sessions, this study sought to determine whether these imaging techniques can reliably detect bone mineral loss and predict skeletal events. If so, our findings might pave the way for more effective and accessible osteoporosis screening protocols, ultimately improving patient outcomes and reducing the burden of fractures in this vulnerable population.

This study was conducted in compliance with the principles outlined in the "Declaration of Helsinki". The institutional Ethics Committee approved the study protocol (Ankara University Ethics Committee, Study Number: 2024000041-1). Given the retrospective nature of the study and the use of anonymized data, the requirement for informed consent was waived by the Ethics Committee. All patient data were anonymized and securely stored in an electronic database to ensure confidentiality.

Patient selection. Patients diagnosed with prostate cancer and followed-up at our tertiary cancer center between 1 July 2006 and 31 December 2023 were screened for eligibility. The data of patients who received ADT and were over 18 years of age were retrieved from the hospital medical recording system. Data, such as age, sex, Gleason score, Eastern Cooperative Oncology Group (ECOG) performance status, treatment protocols, and use of bone-modifying agents, were recorded. Three cross-sectional images of CT or PET/CT performed at different times for other reasons were evaluated by radiologists in terms of osteoporosis, and the mean attenuation of the L5 vertebral body was determined. Skeletal-related events (including pathological fractures and surgery to bone) were identified through a comprehensive review of patients' medical records in the hospital information system.

Statistical analysis. Statistical analysis was performed with R version 4.1 (10). The distribution of the variables was examined using visual (histogram and probability graphs) and analytical methods (Kolmogorov-Smirnov/Shapiro-Wilk tests). Descriptive statistics are given as the mean and standard deviation for normally distributed variables, and as medians with interquartile range for non-normally distributed and ordinal variables. Friedman test was conducted to determine whether there was a significant change in bone density. The Wilcoxon test was performed to test the significance of pairwise differences using Bonferroni correction to adjust for multiple comparisons. An overall 5% type-1 error level was used to infer statistical significance.

A total of 66 patients were included in this study, with a median age of 64 years. The median follow-up period was 45.2 months. Among these patients, the Gleason scores were as follows: 3 patients (5%) had a score of 6, 12 (20.4%) had a score of 7, 11 (18.6%) had a score of 8, 26 (44.1%) had a score of 9, and 10 (11.9%) had a score of 10. Regarding performance status, 61 patients (92.4%) were classified as ECOG 0-1, 4 (6.1%) as ECOG 2, and 1 (1.5%) as ECOG 3.

At the time of diagnosis, 38 patients (57.6%) presented with metastatic disease. Primary treatments administered before relapse included surgery in 13 patients (59.1%) and radiotherapy in 9 (40.1%). The median time to relapse was 44.3 months. Thirty-one patients (47%) received adjuvant ADT, while 35 (53%) did not. Metastasis evaluation revealed that 57 patients (86.4%) had bone metastasis, and 44 patients (66.7%) had visceral metastasis. First-line systemic treatments included docetaxel in 55 patients (83.3%), abiraterone in 5 (7.6%), and enzalutamide in 4 (6.1%). For second-line treatments, 29 patients (47.5%) were treated with enzalutamide, 24 (39.3%) with abiraterone, 4 (6.6%) with docetaxel, 3 (4.9%) with cabazitaxel, and 1 (1.7%) with radionuclide therapy. A total of 60 patients (90.3%) received novel ADT. Bone-modifying agents included zoledronate in 49 patients (98%) and denosumab in 1 patient (2%). Skeletal events were observed in 15 patients (26.2%), and 43 patients (65.2%) underwent bone radiotherapy. Baseline characteristics are summarized in Table I.

The mean attenuation of the L5 vertebral body was recorded as HU in three consecutive imaging sessions, performed at different times for reasons other than osteoporosis. The median HU for L5 attenuation at the first measurement was 131, at the second it was 120, and at the third it was 111.5. The decrease over time was significant (Friedman test, p<0.001). Post-hoc analysis was conducted with the Wilcoxon test, and it was seen that the significance was attributed to the difference between the first two measurements (p=0.01) (Table II).

Average changes in HU values in patient subgroups were evaluated. It was observed that median change considering the whole patient cohort was -4.5; in patients with relapse it was -7.5; in those receiving adjuvant ADT it was -5.7; in those with bone metastasis it was -3.9; in patients with visceral metastasis it was -1.5; in patients receiving novel ADT as first- and second-line therapies, it was 0.4 and -5.5, respectively; in patients who had skeletal events it was -2.8; and in patients receiving bone-modifying agents it was -1.65. No significant difference was observed between patient subgroups (Table III).

In order to examine the effect of changes in bone mineral density on risk of skeletal events, logistic regression with backward method was performed and no significant models were detected. Bayesian regression analyses showed no relationship between skeletal events and changes in bone densities (Bayesian Factor 01: 2.494-2.892, low causality).

Our primary finding indicates that while CT imaging can effectively detect bone mineral loss in patients with prostate cancer undergoing ADT, it does not reliably predict the occurrence of skeletal events. Despite detecting a significant decrease in bone density over time, as measured by HU across consecutive imaging scans, these changes did not correlate with the incidence of fractures or other skeletal complications. This suggests that while CT scans are useful for monitoring bone density, they are insufficient as standalone predictors of skeletal events in this patient population.

This study uniquely employed three consecutive imaging sessions to assess changes in bone density and their association with skeletal events. While CT scans can detect bone mineral loss, they lack the ability to reliably predict these events or identify high-risk fracture subgroups. This highlights the need for novel imaging methods and improved diagnostic tools, such as refined nomograms.

Osteoporosis and skeletal events are an important cause of mortality and morbidity in patients with prostate cancer (11), and the frequency and methods of osteoporosis screening in this patient population are not established issues. In other studies, it has been shown that quantitative CT analysis can be used in the diagnosis of osteoporosis in CT images taken for another reason and is strongly correlated with DEXA findings (12,13).

There appear to be many reasons why skeletal events and pathological fractures are difficult to predict in this patient population. It is known that routine osteoporosis screenings are rarely performed and often not sufficient (4,5). Apart from risk factors such as advanced age and use of ADTs (4,11), patients with bone metastasis, especially those who received bone radiotherapy, make the use of imaging modalities such as X-ray, DEXA and CT to determine bone mineral density challenging. Moreover, in a recent study, although the risk of osteoporosis and 10-year fracture was low, at the start of ADT, one-third of patients had a high prevalence of vertebral fractures (14), and it is known that these fractures can affect the evaluation of bone mineral density. In our study, there was a steady decline in bone mineral density over time, however, this finding was not sufficient to predict skeletal events, which may be related to the reasons mentioned above. In addition, it is known that many bone metastasis are osteoblastic in prostate cancer, leading to increased bone formation, further complicating the evaluation of bone mineral density with imaging techniques.

It is also notable that bone-modifying agents did not affect bone density loss, as measured by CT scans, and bone density loss was similar in patients with skeletal events compared to other patient subgroups. Our findings suggest that imaging techniques alone are not sufficient to predict skeletal events and fragility fractures in this patient population, and other modalities should be used in combination with these. There are several risk-assessments tools and nomograms to predict skeletal events in this patient population. A recent, large-scale study evaluating the Fracture Risk Assessment Tool (FRAX) showed that this method could reliably predict possible fractures in prostate cancer patients (15). Additionally, there are studies showing that the FRAX tool may be useful in risk categorization in patients about to start taking ADT (16-18). Apart from this, new nomograms are being developed for the diagnosis of osteoporosis. In a cross-sectional study including 596 patients, a nomogram was created using age, body mass index, hemoglobin, vitamin D3, testosterone and ADT duration, and its effectiveness in predicting osteoporosis in patients with prostate cancer was demonstrated (19). In another study, an evidence-based algorithm was developed for patients with metastatic prostate cancer, and it was shown that this algorithm improved patient management (20). Novel imaging modalities are also under investigation, including calcaneus ultrasound (7), as well as new magnetic resonance imaging techniques (9) and increased availability of [¹⁸F]-prostate-specific membrane antigen (PSMA)-1007 PET scan findings for diagnosis of osteoporosis (8). While PSMA PET/CT has shown impressive diagnostic accuracy in nodal staging of prostate cancer, its utility varies with disease grade and histology (21), highlighting the continued need for comprehensive assessment approaches. Developing new imaging methods and investigating their combined use with nomograms is deemed necessary to predict bone events in this patient population, and today, more studies on this subject are required.

Our study has several limitations inherent to its retrospective design, which is prone to data incompleteness. A significant limitation was the inconsistent use of DEXA scans during follow-up, which prevented their inclusion in our analysis. Furthermore, in a small subset of patients, it was not feasible to categorize skeletal events by specific sites (e.g., vertebral vs. femoral fractures) nor to differentiate between metastasis-related and osteoporosis-related events. Despite these challenges, our investigation offers a unique contribution to the literature by longitudinally assessing changes in bone mineral density through repeated CT scans and exploring their association with skeletal events, patient characteristics, and treatment modalities. This study represents an important addition to the understanding of bone health management in patients with prostate cancer.

Our study demonstrates that while CT imaging is effective in detecting bone loss in patients with prostate cancer undergoing ADT, it is inadequate for predicting future fractures. This underscores the need for more advanced diagnostic strategies. Sole reliance on CT imaging may miss high-risk patients, indicating the potential value of exploring novel imaging modalities, including PSMA-PET scans, which may offer more precise assessments of bone health in this population. Integrating such advanced imaging techniques with clinical nomograms may significantly enhance risk assessment and patient management. Future research should prioritize the validation of these imaging methods, investigate their combined use with nomograms, and develop a personalized framework for screening and intervention. Such advancements, particularly with the inclusion of PSMA-PET, may lead to better fracture prevention, improved patient outcomes, and the potential reshaping of osteoporosis management guidelines in prostate cancer care.

The Authors have no conflicts of interest to declare.

MY: Conceptualization, writing; ECE: formal analysis, data curation; SCY: data curation, review and editing; SBE: data curation, formal analysis; BG: data curation, review and editing; EY: writing, review and editing; YÜ: conceptualization, supervision, review and editing.

No artificial intelligence (AI) tools, including large language models or machine-learning software, were used in the preparation, analysis, or presentation of this manuscript.