Bone is a common site of secondary tumor deposits because, in addition to its rigid, calcified, outer cortex, it has a richly vascular inner marrow of bony trabeculae, stroma, haematopoeitic tissue and fat (1). Within bone, it is the crucial balance between osteoblastic and osteoclastic elements that maintains its functional strength and rigidity. Metastatic deposits elicit differential responses from the osteblastic and osteoclastic components, which vary with the type of malignancy and result in strikingly different appearances on imaging (2). In some cases, the tumor incites a predominantly osteoblastic response with a resulting increase in calcified sclerotic matrix, as in prostate and breast cancer (3). In other tumor types, the metastasis causes bony destruction (osteoclastic response) without exciting an osteoblastic response, so that metastases (e.g. kidneys, thyroid, lungs) appear lytic and expansile (3). Finally, the tumor cells can simply invade the marrow without influence on the mineral content of the bone (i.e. radio-occult metastases). In many instances there is a mixture of sclerotic, lytic and radio-occult types. As treatment response is often accompanied by an increase in bony sclerosis (“flare response”), it can be difficult to differentiate it from an osteoblastic response to the tumor itself (4). Moreover, once deformed by the presence of metastases, the rigid form of the bony skeleton does not usually remodel sufficiently after treatment to distinguish untreated from treated tumor. Therefore, on morphological imaging, especially X-ray based, evaluation of response to treatment of bone metastases remains difficult.

RECIST were presented more than 2 decades ago and rely principally on unidimensional size measurements (5). Nowadays, RECIST forms the mainstay of response evaluation of solid tumors to treatment and is universally used in clinical trials of solid tumors. Index lesions with well-defined margins, discernable from adjacent parenchyma are required for reproducible measurements, and specific modifications are set out for some tissues (short-axis measurements for lymph nodes, bi-dimensional measurements for brain lesions). However, because of the blastic response of bone to tumor or to treatment, and of the rigid nature of calcified bone where deformity of the cortex persists after treatment, bone lesions were considered unmeasurable by RECIST. Modifications to RECIST (i.e., RECIST 1.1) stated that bone metastases with soft tissue masses > 10 mm could be considered measurable index lesions (6). Nevertheless, as reduction of the soft tissue component renders the lesions unmeasurable by these criteria again, there remains a critical unmet need for a means of quantifying bone lesions and their response to treatment.

The advent of imaging modalities providing information about tissue microstructure or its metabolism has accelerated the identification of skeletal metastases. ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG) PET/CT identifies secondary deposits within bone because of their increased glucose turnover. Its whole-body coverage and increasingly widespread availability has made it of primary importance in cancer staging, particularly in patients where the tumor pathology or molecular profile indicates a high metastatic risk (7, 8). Additionally, techniques such as WB-MRI with DWI have a high sensitivity for identifying highly cellular lesions such as tumors and have been incorporated routinely into the staging of some tumor types such as myeloma (9–11). Dynamic contrast-enhanced MRI (DCE-MRI) for quantitatively assessing vascularization within bone marrow in patients with multiple myeloma was found to be of prognostic significance for these patients (12, 13). While these techniques have their own limitations, they are not hampered by what makes bone lesions unmeasurable by RECIST 1.1 (i.e. radio-occult appearance, sclerotic response and persistent bone deformity on healing). The purpose of this manuscript is to review the MRI and PET techniques available for measuring bone metastases, their opportunities and challenges, and their applicability in various tumor types.

At present, the incidence of bone metastases is 65-75% in advanced metastatic breast cancer, 65-75% in prostate cancer, 60% in thyroid cancer, 30-40% in lung cancer, 40% in bladder cancer, 20-25% in renal cell carcinoma and 14-45% in melanoma (14).

Bone metastases can be classified as osteolytic, osteoblastic, radio-occult, or as a mixed type. Osteolytic metastases are characterized by destruction of normal bone and osteoblastic/sclerotic metastases are characterized by deposition of new bone. Radio-occult lesions have no impact on the mineral content of the bone. Osteolytic lesions are predominantly present in multiple myeloma, renal cell carcinoma, melanoma, non-small cell lung cancer (NSCLC), non-Hodgkin lymphoma (NHL), thyroid cancer, Langerhans-cell histiocytosis and breast cancer, and osteoblastic lesions are present in prostate cancer, neuroendocrine tumors, small-cell lung cancer (SCLC), Hodgkin lymphoma and medulloblastoma (14). Mixed lesions can be found in gastrointestinal cancers and squamous cancers, and 15-20% of bone metastases of breast cancer can be either osteoblastic or mixed (14). Radio-occult lesions can be present in virtually all tumor types. The mechanisms responsible for the impact of metastatic tumor growth on the mineral content of the skeleton are complex and involve the stimulation of osteoclasts and osteoblasts by tumor cells expressing factors. The resulting imbalance between resorption and production of bone matrix subsequently leads to osteoclastic, osteoblastic, or mixed metastatic disease (2).

In osteolytic lesions, bone destruction is primarily mediated by osteoclasts and, in later stages, ischemia can play a role due to the compression of the vasculature (15). Parathyroid hormone-related peptide (PTHrP) induces osteoblasts to produce a receptor activator of nuclear factor κB ligand, which stimulates osteoclast maturation, and thereby plays a critical role in the development of osteolytic lesions. Increased osteoclast activity leads to bone resorption that exceeds the reparative ability of osteoblasts (16). It releases factors from the bone matrix that stimulate PTHrP, thereby creating a vicious cycle. In osteoblastic lesions, osteoblast generation is influenced by transforming growth factor, bone morphogenic proteins (BMP), and endothelin-1 (17). Tumor-derived growth factors stimulate primarily osteoblasts rather than osteoclasts, resulting in deposition of excess abnormal bone. PTHrP can be cleaved by prostate-specific antigen (PSA), resulting in an osteoblastic reaction and decreased bone reabsorption. Furthermore, osteoblast differentiation is influenced by core binding factor alphal, also known as Runx-2 (14). Osteoblast activity may also increase as a reparative process in successfully treated bone metastases, which can be visible on molecular imaging as the so-called “flare phenomenon” and can cause lesions to become denser on radiographs or CT scans (18).

After the tumor cells have left the primary tumor and are in circulation, the bone tumor microenvironment needs to provide a fertile ground (the soil), for the survival and growth of metastatic cancer cells (the seed) (19). Vascular adhesion and extravasation need to occur, and the tumor cells have to remain at the metastatic site. Subsequently, chemo-attractive and adhesion molecules play an important role in the retention of the tumor cells in the bone marrow vasculature. In turn, tumor cells use equivalent molecules, such as chemokines, integrins, osteopontin, bone sialoprotein and type I collagen for organ colonization (20). The microenvironment supports cancer cell survival and growth by producing promoting factors that may contribute to bone metastases development. Subsequently, epithelial-mesenchymal transition occurs, which enables epithelial cells to migrate to a new environment. While this occurs mainly during embryogenesis, in cancer cells this process denotes the invasive phenotype (21).

Sex-associated differences exist in bone metastasis formation from breast-, lung- and prostate cancer. In breast cancer, estrogen influences the bone microenvironment by creating and conditioning a favorable niche for colonization of breast cancer cells. Patients with estrogen receptor α positive (ER+) tumors have bone metastases three times more often than do patients with ER- tumors (22). In lung cancer, it is reported that females more often have bone metastases due to a more favorable bone microenvironment for metastasis formation. In prostate cancer patients, a decrease in the androgen-to-estrogen balance results in bone metastasis formation, with a potentially important role for ERβ that may be similar to that in breast cancer. Androgens as well as estrogens have an influence on osteoblast proliferation and on bone resorbing osteoclasts. In both males and females, estrogens have a dominant effect on bone maintenance and can directly inhibit osteoclasts. Furthermore, androgens directly contribute to male periosteal bone expansion, mineralization, and trabecular bone maintenance (23).

The time from primary diagnosis to the development of bone metastasis can range from months to decades. This implies that tumor cells can lay dormant for significant periods of time after they leave the primary site. It has been shown that the bone is an important reservoir for dormant tumor cells. The best-illustrated cases for clinical dormancy are in breast cancer, where ER+ patients show late recurrences, sometimes decades after removal of the primary tumor. Latent bone metastasis formation likely depends on estrogen regulation, and it is significantly higher in ER+ cases (24).

Bone metastases have unique disease-specific characteristics, such as longevity, fracture healing rates, local and systemic disease progression, and sensitivity to adjuvant treatments. Bone metastases from lung cancer and renal cancer can also show acral distribution (25). Patients with bone metastases of lung cancer historically showed a median survival of approximately 6 months (14). Treatment options for patients with identifiable mutations include immunotherapy and epidermal growth factor receptor tyrosine kinase inhibitors, with evidently improved survival benefit (26). Bone metastases of lung cancer are, in general, sensitive to radiation therapy (27).

The median survival of breast cancer patients with bone-only metastasis is 36 months (28). The medical treatment of breast cancer depends on the hormone receptor and HER-2/neu status and is different for premenopausal and postmenopausal women (25). Furthermore, pain reduction can be achieved, and skeletal-related events and the development of new skeletal lesions can be prevented by the use of bisphosphonates or denosumab, due to their ability to limit bone resorption. Bone metastases of breast cancer are in general radiosensitive, resulting in a lower proportion of surgical treatments (29).

Men with prostate cancer, a good performance status, and bone-only disease have a median duration of disease control after androgen blockade of 4 years and a median survival of 53 months (14). Bone metastases of prostate cancer have a predilection for the axial skeleton, resulting in an increased risk for spinal cord compression (25). However, due to the osteoblastic nature of the metastases, skeletal-related events are relatively uncommon. Also, bone metastases of prostate cancer tend to be radiosensitive, which allows a higher proportion of nonsurgical treatment. In case of a pathologic fracture, healing rates are higher than for most other metastatic carcinomas (29). Treatment with Radium-223, a calcium-mimetic and alpha-emitter that selectively binds to areas of increased bone turnover, results in significantly prolonged OS in patients who had castration-resistant prostate cancer and bone metastases (30).

Quantitative imaging of bone metastases beyond morphology has been studied in preclinical studies on the functional and molecular level using MRI and PET. In these studies, quantitative biomarkers in skeletal lesions were assessed and validated with the underlying histology. Thereby, DCE-MRI parameters in bone metastatic lesions from breast cancer associated with blood volume and vessel permeability were correlated with vessel maturity, while the apparent diffusion coefficient (ADC) from DWI was associated with tumor cellularity as assessed by cell nuclei staining (31). Treatment monitoring in an animal model of osteolytic breast cancer could be performed reliably using DCE-MRI and ¹⁸F-FDG PET, while therapy response could be detected through functional and metabolic techniques earlier than through morphological imaging (32, 33). Integration of parameters from DCE-MRI and ¹⁸F-FDG PET by machine learning algorithms enabled the detection of early pathologic processes in the bone marrow preceding morphologic changes in bone structure (34). Thus, parameters from functional and metabolic MR and PET imaging are powerful tools to quantify pathophysiologic processes during colonization of bone marrow and to determine response to treatment of skeletal metastasis.

On the molecular level, PET is the method of choice to determine molecular structures expressed in bone metastases, such as integrins alpha_vbeta_3/5 or the chemokine receptor CXCR4 (35, 36). Although a major limitation of MRI is the lack of sensitivity when compared to PET, a strategy of signal amplification using a pair of enzymes and an appropriate reducing substrate was presented recently to non-invasively assess epidermal growth factor receptor (EGFR) expression in MRI (37). Besides MRI and PET, other imaging modalities may also be used to determine molecular information in bone metastases, such as ultrasound with its high spatial resolution and unique contrast characteristics of gas-filled microbubbles for enabling the assessment of intra-vascular targets such as vascular endothelial growth factor receptor-2 (VEGFR-2) expressed in bone metastases (38). Thus, molecular imaging strategies for molecular characterization of skeletal lesions have been developed for PET but also for MRI and ultrasound, which are suitable for clinical translation in the near future.

The unique potential of hybrid imaging, as reviewed by Schmidkonz and colleagues, lies in the assessment of complementary parameters on the morphologic, functional, metabolic and molecular levels of bone metastases from different modalities in a single examination (162). When combining PET with CT in a PET/CT study, the CT component enables assessment of bone morphology and osseous destruction, while MRI in a PET/MRI hybrid study will offer superior soft tissue contrast. Due to the (still) novelty and increased cost and complexity of PET/MRI, this technique currently is primarily compared to PET/CT for assessing the respective potential of these two imaging approaches for evaluating bone metastases.

When comparing the performance of ¹⁸F-FDG PET/CT with ¹⁸F-FDG PET/MRI for the assessment of malignant bone lesions, the overall performance of PET/MRI has been found to be equivalent to PET/CT for the detection and characterization of bone lesions when these hybrid techniques were performed sequentially (163). However, in PET/MRI, lesion delineation and allocation of PET-positive findings were found to be superior to PET/CT (163). Samarin and colleagues reported similar results from a comparison of ¹⁸F-FDG PET/CT with ¹⁸F-FDG PET/MRI in 24 patients with bone metastases from different primary tumors (164). The overall detection rate was not significantly different between PET/CT and PET/MRI, but the latter provided higher reader confidence and improved conspicuity as compared with PET/CT (164).

In a prospective comparison of the diagnostic accuracy of ¹⁸F-FDG PET/MRI and CT, PET/MRI was significantly better than CT for the detection of bone metastases in patients with newly diagnosed breast cancer (165). Also, in a particular series of 109 breast cancer patients, PET/MR demonstrated an improved sensitivity over ¹⁸F-FDG PET/CT alone, where the sensitivity of PET/MR and PET/CT were 96% and 85%, respectively (166). In men with biochemical recurrence of prostate cancer following curative therapy, ⁶⁸Ga-PSMA-11 PET/MRI demonstrated a high detection rate especially for recurrent disease with low PSA values, but included all sites of local or distant recurrence including lymph nodes and bone (167). In 26 patients with prostate cancer, ⁶⁸Ga-PSMA-11 PET/MRI and PET/CT performed equally regarding the PET component for detection of bone metastases, while two PET-positive skeletal metastases could be confirmed on contrast MRI, but not on CT (168).

An interesting approach for patients with both osteolytic and osteoblastic metastases from breast or prostate cancer was proposed by Sonni and colleagues (169). Combining ¹⁸F-FDG and Na¹⁸F in PET/MRI was superior for the detection of skeletal metastases as compared to whole-body bone scintigraphy (169). This approach includes in an innovative manner both a radiotracer (Na¹⁸F) for the assessment of primarily osteoblastic activity in osteoblastic lesions, and another tracer (¹⁸F-FDG) for assessing increased glucose metabolism in the soft tissue component of predominantly osteolytic metastases. Based on the data and results referenced above, PET/MRI appears rather superior to PET/CT for the detection of metastatic bone lesions, but it lacks the morphologic information of bone and osteoblastic bone formation derived from the CT component, which might be mitigated some through innovative approaches as reported by Sonni and co-workers.

In 2009, the PET Response Criteria in Solid Tumors (PERCIST) were introduced for ¹⁸F-FDG PET (77). Later on, along with the detailed describing of the ¹⁸F-FDG PET requirements to allow quantitative expression of the changes in PET measurements and assessment of overall treatment response, a Simplified Guide to PERCIST 1.0 was published (170). The PERCIST criteria enable avid bone target lesions to be selected based on their metabolic activity, and response to be measured objectively based on the changes in metabolic activity even in the absence of an evident anatomic change. PERCIST, however, only considers the change in uptake of a single target lesion when assessing response, which is the lesion with the highest SUV_peak value normalized for lean body mass (SUL_peak). New lesions, in the bone or elsewhere, result in progressive disease by definition. The target lesion may or may not be within the bone, but all bone lesions have to be considered in the selection of target lesion. Of note, there is no impact of changes in volume of lesions, only the uptake concentration is considered. Compared to RECIST, PERCIST represents a major step forward for bone assessment as it considers bone lesions equally to any other lesions anywhere else in the body.

The PERCIST approach focusses on the remaining hottest lesion and has similarities with the therapy response criteria for lymphoma, where the most active remaining lesions play a dominant role (171). This “hottest lesion” centric approach is well tailored to therapies with curative intent, but it might miss a beneficial effect in non-curative therapies, where tumor control and tumor bulk reduction are clinically relevant achievements. A recent approach in image analysis is about abandoning the selection of target lesions, as determined on baseline or post-therapy scans, and aiming to take the entire tumor bulk in consideration. The high contrast of modern oncological PET tracers [e.g., ¹⁸F-FDG, somatostatin receptor (SSTR) and PSMA ligands, ¹⁸F-DOPA, ¹⁸F-MFBG (172)] permits straightforward three-dimensional segmentation of lesions; by segmenting all lesions, the total tumor burden can then be obtained. This type of analysis does not distinguish between bone and non-bone lesions and thus puts bone metastases on par with other metastases.

There is evidence that the baseline metabolic tumor volume (MTV) is an important prognostic factor, e.g. in NHL, NSCLC and multiple myeloma, as well as in prostate cancer patients treated with the bone-seeking agent radium-223 (173–176). Furthermore, MTV can be combined with metrics of tumor distance within a patient to not only represent volume, but also dissemination for better reflecting prognosis, as shown in NHL (175). Evaluation of the changes in MTV and/or TLG have been shown to outperform PERCIST-based approaches in tumors with frequent bone lesions, such as Ewing sarcoma and osteosarcoma (177–180). Volumetric determination on PET is not hampered by bone/soft tissue interfaces, taking the total tumor burden into account in combination with the metabolic activity. Total tumor burden analyses can be combined with specific organ segmentation either based on PET or CT, e.g. for spleen (for lymphoma) and bone, to generate organ-specific tumor burden (181, 182). Furthermore, the segmentation leading to total tumor burden or organ-specific tumor burden can also be used as a mask to determine specific radiomic features, which can provide even more information for response evaluation (180).

Although promising, some challenges remain to the application of tumoral volumes for routine therapy response monitoring: (i) lack of standardization of uptake thresholds for PET-positive tumor delineation; (ii) still too time consuming for clinical routine; (iii) no prospectively defined response criteria. Especially regarding i and ii, it is expected that advances in tumor segmentation, e.g. with further automatization of the segmentation process and contributions from deep learning-based AI algorithms, will increase the robustness, accuracy and feasibility of total tumor segmentation to routine clinical practice levels (141, 183). Based on results from current and ongoing studies, automated tumor segmentation should actually be one of the key expected improvements from AI applications to PET (and other) imaging (160).

At any rate, similar analyses can be developed for non-FDG tracers, e.g. with SSTR ligands in neuroendocrine tumors and PSMA ligands in prostate cancer, Na¹⁸F in breast and prostate cancer patients (184–187). For evaluation of response on PSMA PET, consensus criteria have recently been proposed with specific cut-off values for both uptake and volume (188).

A specific case of total tumor burden imaging that is worthy to mention is the use of the bone scan index (BSI), which is a metric based on 2D planar bone scintigraphy that reflects the fraction of bone showing increased turnover due to metastatic invasion (189). It has been proposed two decades ago as a metric for tumor burden and response assessment in metastatic prostate cancer (189). Changes in BSI under treatment have been shown to correlate with OS in patients with metastatic castrate-resistant prostate cancer (mCRPC) treated with a range of therapies (190). This has been corroborated in multicenter trials in mCRCP patients treated with abiraterone acetate and with radium-223 dichloride (191, 192). Although promising, it is expected that the shift from 2D to 3D imaging and the increasing use of novel PET tracers that can pick up lesions outside of the bone as well as bone lesions (e.g. PSMA ligands) will eventually displace the currently widespread adoption of BSI for therapy response. Accordingly, similar but PET-based metrics from PSMA and/or Na¹⁸F PET will likely outperform and replace BSI.

Modern imaging with PET and MRI allows bone metastases to be detected and assessed both before and after therapy, without the drawbacks of X-ray based imaging techniques (e.g. radiographs, CT). These techniques assess bone metastases within the same framework, as metastases in other organs. They further allow total tumor burden to be assessed within a single imaging session, and also the development of response criteria that include the bone, thus filling a critical gap in the RECIST1.1 framework. The EORTC, PERCIST and recent PSMA PET criteria are examples of criteria that take bone metastases in consideration, on-par with extra-osseous lesions. PET and/or MRI can detect and characterize bone metastases of various types (e.g. lytic, sclerotic, radio-occult or mixed) independently from the bone density changes. In contrast with CT, they are not affected by changes in bone mineralization induced by the tumor(s), and are not dependent on soft-tissue components (as required by RECIST 1.1). Whole body MRI including modern techniques such as DWI, DCE-MRI and mpMRI can provide both detailed information on anatomical structures as well as functional information on individual lesions and whole body tumor burden. Modern PET imaging is performed on hybrid cameras, with CT (from PET/CT) allowing assessment of the bone mineral content (including fractures), while MRI (from PET/MRI) can more often clarify a correlate for the lesions observed on PET. Total tumor burden, incorporating bone metastases on par with other metastases, is an attractive approach to be applied in most PET tracers. While advances in algorithms and deep-learning contributions are expected to permit the determination of total tumor burden metrics in actual clinical routine before and after therapy, response criteria through total tumor burden assessment are currently developed, taking into consideration the tracer, therapy and underlying cancer type.

All authors listed have made a substantial, direct, and intellectual contribution to the work, and approved it for publication.

KH reports personal fees from Bayer, personal fees and other from Sofie Biosciences, personal fees from SIRTEX, non-financial support from ABX, personal fees from Adacap, personal fees from Curium, personal fees from Endocyte, grants and personal fees from BTG, personal fees from IPSEN, personal fees from Siemens Healthineers, personal fees from GE Healthcare, personal fees from Amgen, personal fees from Novartis, personal fees from ymabs, all outside the submitted work. CD reports consultancy for Sirtex, Terumo and PSI CRO, speaker fees from Terumo and Advanced Accelerator Applications and is a member of advisory board for Terumo and Ipsen. EL reports grants from Fondazione AIRC and Italian Ministry of Health, royalties from Springer, lecturer fees from MI&T congressi and ESMIT. LF reports speaker fees from Sanofi, Novartis, Jannssen, and General Electric, congress sponsorship from Guerbet, industrial grant on radiomics from Invectys and Novartis, and co-investigator in grant with Philips, Ariana Pharma, Evolucare.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.