Assessment of acute bone loading in humans using [¹⁸F]NaF PET/MRI

Acute loading of bone tissue is thought to stimulate bone formation and is of growing interest clinically and in the study of bone physiology. Abnormal bone physiology is not only a key element in joint disease and osteoporosis, but skeletal fragility is directly related to mortality [1‐3] and risk of fracture [4]. Studies of bone adaptation to loading have shown that stress on bone cells and strain-mediated fluid flow are crucial in regulating bone metabolism [5‐11]. However, the acute response of loading in bone is still poorly understood and difficult to measure in humans in vivo.

Molecular information from PET has shown promise in early detection of metabolic abnormalities of bone metabolism in osteoarthritis [12, 13], associations to bone pain [12‐16], and early indication of bone degradation in diseases such as osteoarthritis [14] and osteoporosis [17]. PET has also been used to demonstrate changes in glucose uptake and blood flow in bone marrow in response to exercise loading [18‐20]. [¹⁸F]-sodium fluoride ([¹⁸F]-NaF) is a well-established bone-seeking agent which may serve as a marker to study bone turnover. In particular, kinetic modeling of dynamic [¹⁸F]NaF uptake can quantify bone physiology including bone perfusion (K₁), bone mineralization (k₃), and tracer plasma clearance (K_i). Animal studies [8, 21‐24] have shown both an acute hyperemia [22, 24] and a large increase in [¹⁸F]-NaF standard uptake values (SUV) in response to acute loading, which lasts up to 7 days after loading [21, 22, 24]. Furthermore, this response has been positively correlated with the force intensity of the applied load on the bone [21]. The dynamics of [¹⁸F]NaF uptake after acute bone loading have not been reported in humans. To date, human studies of bone adaptation to loading have been based on changes in structure and mineral density over periods of weeks and months of high-impact exercise. The aim of this study is to evaluate quantitative measures of [¹⁸F]NaF uptake and tracer kinetics to assess the acute physiological vascular and metabolic response of bone to loading in the human knee using PET/MRI.

Exercise induced a response in [¹⁸F]NaF kinetics in all bone tissues (Fig. 2 and Fig. 3), with increases in SUV, SUVmax, K_i, blood volume, and k₂, along with a decrease in extraction fraction in all bone tissues (p <  0.001). In cortical bone, k₃ increased after exercise as opposed to a post-exercise decrease of k₃ in trabecular and subchondral bone (p < 0.01). SUV increases were lowest in cortical bone (47%) and highest in subchondral bone (131%) (p < 0.001). SUV increases were significantly higher in the step-up (right) knee (p < 0.05). Absolute and relative changes in SUV values and kinetic parameters for cortical, subchondral, and trabecular bone are presented in Table 1. On an individual level, regional variations in SUV increases (Fig. 3) and changes in kinetic parameters (Fig. 4) were observed between medial–lateral compartments as well as between step-up (right) and drop–land (left) legs.

However, on a group level, significant differences in increases were only observed in SUV, K_i, K₁, and blood volume of subchondral bone between the step-up leg (p < 0.05) and the drop–land leg (Fig. 5). All compartments of subchondral bone within each knee showed significant (p < 0.05) post-exercise changes of SUV and all kinetic parameters (Fig. 6) with the exception of blood volume and k₂. The patella had the highest absolute changes in SUV and kinetic parameters (p < 0.05) of all the subchondral ROIs, as well as the largest differences between the step-up and drop–land legs (Fig. 6).

Higher post-exercise K_i^NLR values were measured in all tissues. This was driven by increases in K₁ values that exceeded the decreases in extraction fraction. An increase in fluoride flux back to blood plasma (k₂) in all tissues contributed to a decrease of overall extraction fraction (k₃/(k₂ + k₃)). The decrease in extraction fraction was found to be inversely proportional to K₁ increases (R² = 0.43, p < 0.01) (Fig. 7).

K_i and SUV values were highly correlated (R² > 0.8) both before and after exercise, although the correlation coefficient was altered due to a relatively larger increase in SUV values than K_i^NLR values. Prior to exercise, the regression slope of K_i^NLR against SUV was 85, whereas after exercise the slope increased to 105 (Fig. 7). The correlation between K_i^NLR and K_i^pat remained unchanged in both conditions (Fig. 8).

Increases in [¹⁸F]NaF uptake parameters did not always follow a regional or compartmental pattern. Eight ROI_focal were identified in the post-exercise [¹⁸F]NaF PET images of six subjects (Fig. 9). Of these, three of the focal points already had significantly higher activity (Fig. 9c) than the surrounding subchondral tissue at baseline, and five were unremarkable at baseline (Fig. 8a, b). The relative mean increase in SUV was 342% (209–518%) for the focal points appearing normal at baseline, and 111% (82–180%) for the focal points which were identifiable at baseline .

Differences between venous blood samples and IDIF values measured at 50 min were on average within 0.2 KBq/ml, which corresponds to a coefficient of variation (CV) of 12%. Comparing the computed IDIF between repeated acquisitions across 12 subjects, a CV of 8.4% was observed at 2 min after injection, 10.6% at 10 min, and 9.9% at 25 min. The input function of the second injection was on average 0.25 KBq/ml higher than the first injection throughout the 50-min time period across all subjects.

In this study, we evaluated kinetic modeling of dynamic [¹⁸F]NaF PET uptake to detect changes in bone physiology after an acute exercise bout. In healthy volunteers, exercise induced a significant change in uptake parameters, suggesting that [¹⁸F]NaF PET may provide a sensitive measure of the metabolic and hemodynamic response of bone to acute joint loading. Notably, the difference in exercise task performed by the right and left leg (step-up vs drop–land) was reflected in quantifiable differences of the patellar acute response between the two knees. Other regional differences were observed in parameter changes throughout the subchondral bone of each knee, suggesting discrepancies in the mechanical loading of these subsections. Further, focal areas of elevated [¹⁸F]NaF PET uptake may identify regions where there is an inconsistent response of the bone–cartilage unit to loading in areas that appear normal on MRI. Some focal points were evident prior to exercise and maintained a similar elevated uptake compared to surrounding tissue post exercise, and others were not evident until post exercise. These findings suggest that kinetic modeling of [¹⁸F]NaF PET may provide a promising non-invasive method to study the effects of acute loading on bone remodeling, which may increase our understanding of the effects of loading on bone and joints in healthy individuals as well as in patients with arthritis/arthrosis.

The changes in kinetic parameters observed included large increases in perfusion, as measured by fluoride influx (K₁), and blood volume which coincided with a decrease in the rate of fluoride extraction k₃/(k₃ + k₂) (Table 1 and supplementary material). This combination is indicative of a hyperemia effect, where an increased delivery of substrates exceeds consumption during a vascular response to stimulation. Changes of blood volume from near zero at baseline to a volume fraction of 1–2% after exercise indicate a vasodilation of the capillary bed as well. In humans, blood flow increases of 130% at the mid femur have been reported for bone, bone marrow included [19], as have increases in blood volume in the tibia [32] during exercise. Further, the exercise-induced hyperemia in bone marrow has been shown to be in part regulated by opposing vasodilators (nitric oxide, prostaglandin, and adenosine) and vasoconstricting factors such as the α-adrenoreceptor. [20, 33]. Although clinical bone-loading studies using [¹⁸F]NaF are sparse, our findings are similar to previous animal studies. Rat forelimbs were found to have increases in the rates of fluoride influx (K₁ by 113%), fluoride metabolism (K_i by 133%), and fluoride incorporation into bone (k₃ by 13%), as well as an increase in bone blood flow (26%) in cortical bone after heavy loading [24]. Discrepancies in the magnitude of change reported previously and this study may be due to differences in type and magnitude of loading, timing of the post-exercise scanning, and/or differences in bone physiology between rats and humans. Another study of [¹⁸F]NaF SUV values in the shaft of the rat forelimb reported an increase of 50% immediately after bone loading with comparable force magnitudes, which is similar to the 47% increase we measured in cortical bone [21]. However, the same study also observed increases in SUV to be greater with higher loading forces and that SUV values continued to increase over time peaking 4–7 days after loading. Time delays of up to 5 days have been registered between a single period of mechanical loading in vivo and observed increases of collagen and mineral apposition on surface of the loaded bone matrix [7]. Similarly, increased blood flow in the rat ulna has been found to peak 5 days after loading [22]. It should be noted that the aforementioned studies used bone-loading devices as opposed to the more physiologic type of exercise in the present study. Such loading studies also tend to employ loads that exceed forces under normal exercise, which means the time and load dependency of a measured response in bone requires more study. Still, the interconnection of vascular fluctuations in bone with an anabolic effect on bone remodelling has been established [34‐36].

The changes in [¹⁸F]NaF kinetics, total uptake, and SUV induced by exercise in this study suggest a localised response in bone tissue that may be load-dependant, with high inter-subject variability. Although the step-up and jump–landing exercises are functionally different, they result in a similar loading environment in the knee joint. The tibiofemoral joint experiences compressive forces from both ground reaction forces and muscle forces, while the patella experiences tensile forces from the quadriceps and compressive forces through its articulation with the trochlea. Stair ascent and descent are similar biomechanical tasks to the step-up and jump–landing exercises. There are slight differences in peak ground reaction forces between stair ascent:(1.1 bodyweights) versus stair descent (1.4 bodyweights), but both involve peak quadriceps forces of 2.4 bodyweights [37]. As a result, the right and left tibiofemoral joints may have experienced similar loading, but the right patella may have transmitted greater quadriceps forces than the left. In this study, both legs showed the highest SUV and K_i^NLR increases in the patella, which may correspond with the large forces transmitted through the patella. The right patella, however, demonstrated higher SUV and K_i^NLR than the left, which may be explained by the greater quadriceps forces in the step-up activity. The whole-body biomechanics of the two tasks are similar, but tissue-level strains have high inter-subject and spatiotemporal variability [37]. Differences in movement dynamics, varus/valgus alignment, muscle strength, and coordination strategies could all contribute to a high degree of inter-subject variability in forces applied on the bones. Even if the applied forces were identical between subjects, local bone strain is also dependent on the geometry and microstructure of each individual’s bones, resulting in further variability. We also observed intersubject discrepancies in the spatial distribution of kinetic/SUV changes, which may reflect inter-subject differences in bone strain. Furthermore, these discrepancies imply that the physiological mechanisms responsible for the altered kinetics are locally regulated, as opposed to a uniform increase in flow to the bones of the knee region.

SUV values increased more than K_i values even though the two were correlated before and after exercise. One reason for this bias could be the 0.25 KBq/ml extra blood activity (an increase of approx 10% during the 30 to 50 min time period) after the second injection. SUV may also be artificially elevated due to a reduction in the extraction of fluoride in other areas of the body, such as the kidneys. Exercise is known to reduce blood flow to other areas of the body, including the kidneys, via activation of the sympathetic nerve system [38]. Increasing the proportion of fluoride available to bone tissue in the knee would not alter K_i, as it is accounted for by the input function. SUV, on the other hand, is calibrated to the injected dose and does not account for remaining activity in the blood from the first injection or for changes in fluoride extraction elsewhere in the body including the kidneys [38].

Findings of focal areas in subchondral bone with abnormally high increases in ¹⁸F-NaF uptake after exercise, ROI_focal, are of particular interest. Many of these regions showed no abnormalities in ¹⁸F-NaF uptake compared to adjacent regions at baseline pre-exercise or structural subchondral bone abnormalities on MRI. This suggests that ¹⁸F-NaF PET may identify regions where there is an improper response of the bone–cartilage unit to acute loading in areas that appear structurally normal on MRI. Biomarkers that can effectively assess early breakdown of joint function, before structural changes are seen, are crucial for the development and evaluation of disease-modifying treatments, and increased ¹⁸F-NaF uptake detected after acute loading may be a promising candidate. Longitudinal studies are necessary to identify the relationship of these regions of increased uptake after acute joint stress and the pathophysiology of bone and cartilage degeneration in these regions.

There are several limitations in this study. First, the study was designed for high loading of the knee joint to ensure detection of an eventual osteogenic response from exercise. The study design is, for this reason, limited in its ability to interpret the relationship between the amount of stress applied to bone tissue and the corresponding response. The exercise paradigm included both high strain magnitudes and high strain rates on the knee joint [7], and used the maximum number of cycles reported to ensure saturation of mechanostimulation [5]. To determine the relationship between the stress applied to bone and the resulting response, all three of these elements would have to be varied in a controlled fashion.

This study used a cross-sectional analysis of healthy volunteers with no symptoms or history of knee injury or pain. Further, no histopathology was acquired in ROIs with abnormal focal increases in uptake after exercise. While increased bone activity detected by ¹⁸F-NaF after acute loading is suggestive of higher bone loading due to a breakdown of proper joint function, longitudinal studies or histopathology are needed to confirm degenerative changes in these areas. A study focusing on patient groups with known pathology would help assess the utility of stimulating bone tissue in diagnostics. The number of subjects in this study is also small, which limits its ability to identify factors such as BMI, varus/valgus alignment, disease, or activity level that could alter the kinetics in bone tissue. The exercise in this study is also inappropriate for many patient groups. A paradigm designed for patients could, however, induce a similar response, given that ~95% of the osteogenic effect can be achieved after climbing only ~20 to 40 steps (~2000 μƐ in compression) [5, 39, 40].

Furthermore, there are technical limitations to consider. Kinetic studies are more complicated than conventional imaging, requiring more computation, expertise, and the robust determination of an input function. The IDIF methods used in this study have previously been shown to correlate with late-stage venous blood measures and to be reproducible. Further, input function values in this study correspond well with literature values for arterial sampling [29], but using arterial sampling is still considered a gold standard. The use of two consecutive injections was accounted for, but may create an adverse bias in the results, as blood activity levels from the second injection were slightly higher than the first. The blood volume measured by this method is difficult to verify independently in humans with a gold standard. Lastly, despite the numerous advantages from combining PET imaging with MRI in knee examinations, PET/MR is also known to underestimate SUV values by approximately 10% due to MR-based attenuation correction of its PET images [41].

The ability to image the physiological response of bone tissue to loading would be a valuable clinical tool. Decreased mechanosensitivity in bone tissue has been identified in osteoporosis [42, 43] and patients with spinal cord injury [32], and to decline in the general population with age [44‐46]. Medicinal treatment or therapies employing mechanostimulation including exercise [47], muscle stimulation [6, 48], vibration therapy [49], or drug treatment [17, 50] could better be imaged using [¹⁸F]NaF PET. For prostate and breast cancer, studies have shown bone-loading exercise to be a safe and viable suppressant of tumor growth in bone metastasis [51, 52]. The use of kinetics, even at rest, haa been proven to be clinically feasible and has the potential to offer more information. For instance, K_i has been shown to be highly reproducible in the human spine [53, 54] and to be more valuable in monitoring osteoporosis patients than SUV values [55]. Subchondral bone changes that are present prior to and during the development of OA have demonstrated increased bone blood flow and bone remodeling in [¹⁸F]NaF PET studies [15]. An association between pain and mean normalized standardized uptake values (SUV) has been established in different bone types [12, 15, 16], and increased [¹⁸F]NaF SUV values may precede visible lesions [12, 14]. Thus, detecting early changes in bone remodeling and bone physiology with [¹⁸F]NaF kinetics could help in understanding disease. In those cases where pathology causing elevated SUV is poorly understood or where there is weak contrast to healthy bone, the response of bone tissue to mechanostimulation could further elucidate dysfunction.

Loading acutely alters bone physiology affecting [¹⁸F]NaF-PET kinetics, with a local response depending on tissue, site, and exercise. Our data support previous evidence of mechanostimulation initiating an early hyperemia phase in bone adaptation. These findings suggest that kinetic modeling of [¹⁸F]NaF PET may provide a non-invasive method to study the effects of acute loading on joint biomechanics and bone remodeling, which may have large implications for our understanding of early stages of arthritis/arthrosis.