The Effects of Exercise and Activity-Based Physical Therapy on Bone after Spinal Cord Injury

Spinal cord injury (SCI) produces paralysis and a unique form of neurogenic disuse osteoporosis that dramatically increases fracture risk at the distal femur and proximal tibia. This bone loss is driven by heightened bone resorption and near-absent bone formation during the acute post-SCI recovery phase and by a more traditional high-turnover osteopenia that emerges more chronically, which is likely influenced by the continual neural impairment and musculoskeletal unloading. These observations have stimulated interest in specialized exercise or activity-based physical therapy (ABPT) modalities (e.g., neuromuscular or functional electrical stimulation cycling, rowing, or resistance training, as well as other standing, walking, or partial weight-bearing interventions) that reload the paralyzed limbs and promote muscle recovery and use-dependent neuroplasticity. However, only sparse and relatively inconsistent evidence supports the ability of these physical rehabilitation regimens to influence bone metabolism or to increase bone mineral density (BMD) at the most fracture-prone sites in persons with severe SCI. This review discusses the pathophysiology and cellular/molecular mechanisms that influence bone loss after SCI, describes studies evaluating bone turnover and BMD responses to ABPTs during acute versus chronic SCI, identifies factors that may impact the bone responses to ABPT, and provides recommendations to optimize ABPTs for bone recovery.


Introduction
An estimated 250,000 to 500,000 new spinal cord injuries (SCI) occur worldwide each year [1], with males representing~80% of the population [2]. Roughly one-third of these are motor-complete SCIs that result in permanent sublesional paralysis, while the remainder are incomplete and retain voluntary contractility in some muscles that are innervated below the lesion [2]. Locomotor dysfunction is the most recognizable symptom of SCI and is accompanied by other medical consequences that develop in this population [3], including severe osteoporosis and high fracture risk [4], which worsen with increasing SCI severity [5][6][7][8][9] and injury duration [10,11].
Bone loss after SCI is termed neurogenic or disuse osteoporosis and is confined to the sublesional skeleton [10][11][12], with the most rapid and prevalent bone deficits occurring at the distal femur and proximal tibia regions [11,13,14]. At these sites, 50-100% lower trabecular bone mineral density (BMD) develops in individuals within the first two to

Mechanisms Regulating Bone Loss after SCI
The molecular mechanisms that propagate the uncoupled bone turnover that is present in the paralyzed limbs after SCI and that drive the exceedingly rapid bone loss in this population, in comparison with other disuse conditions, require further elucidation. Given that disuse is a factor that mediates SCI-induced bone loss, it is likely that osteocytes (primary bone mechanosensor) influence neurogenic osteoporosis. Osteocytes reside within the calcified bone matrix and communicate with other osteocytes and with osteoclasts, osteoblasts, and other cells that reside on bone surfaces via dendritic projections that emerge from the osteocyte cell body to form an interconnected dendritic network. Dendrites provide one means by which the osteocytes sense alterations in localized bone strains that result from disuse or imposed loading and transduce this information to the osteocyte cell body, a process that is referred to as mechanotransduction. In response to this stimulus, osteocytes release a host of nuclear-derived signaling molecules (e.g., receptor activator of NF-κB ligand (RANKL), osteoprotegrin (OPG), sclerostin, and others) that orchestrate osteoclastic and/or osteoblastic bone remodeling. Readers are directed to the following reviews that discuss osteocyte mechanosensors and mechanotransduction-associated signaling pathways [60,61].
While it is likely that osteocytes orchestrate the skeletal responses to SCI, few studies have directly assessed this possibility or described how osteocytes respond to paralysis or to imposed bone loading after SCI. Qin et al. observed osteocyte morphological aberrations in rats within seven-weeks of spinal transition, including reductions in the dendritic length and dendritic number, along with altered osteocyte cell body shape [62]. The molecular mechanisms that regulate these morphologic changes require further elucidation. However, administration of a monoclonal sclerostin antibody that binds and inactivates the circulating sclerostin was shown to preserve the dendritic length and osteocyte cell body morphology in this model [62], suggesting potential autocrine regulation. Regardless, Int. J. Mol. Sci. 2022, 23, 608 5 of 42 pathways [61]. Several negative regulators of Wnt-signaling exist, including sclerostin, which binds the LRP5/6 receptors and prevents Wnt ligands from initiating Wnt-signaling. Sclerostin is known to mediate bone loss that develops in response to unloading [73] and bone gain resulting from imposed skeletal strains, with sclerostin suppression being essential for the osteogenic responses to mechanical loading [74]. Readers are directed to the following review for further discussion on Wnt signaling in bone [61].
Alterations in Wnt signaling accompany bone loss in the paralyzed limbs in rodent SCI models. For example, several proteins that are involved in Wnt signaling (LRP5, Wnt1, and Wnt3) are repressed at the distal femur and proximal tibia within a few weeks of SCI, while the number of osteocytes that stain for sclerostin are increased [75] and SOST (gene encoding sclerostin) is 300% higher. These changes likely influence the 50% lower β-catenin protein expression that has been observed at the proximal tibia after SCI, along with the reduced β-catenin expression that is present in cultured mesenchymal cells that are derived from spinal transected mice. Elevated sclerostin also appears to mediate SCI-induced bone loss, given that SOST-/-mice are protected against bone loss and suppressed osteoblastogenesis after spinal transition [76]. Moreover, a pharmacologic sclerostin antibody that binds and inactivates sclerostin has been shown to completely prevent distal femur and proximal tibia bone loss in rodents when it is delivered immediately after severe incomplete [53] to complete SCI [62] by promoting bone formation, and to reverse trabecular and cortical bone loss when it is delivered after bone loss emerges [77].

Activity-Based Physical Therapy (ABPT) after SCI
Given the role of disuse in mediating the musculoskeletal decline after SCI, a large body of research has focused on ABPTs and other exercise-based regimens that reload the impaired limbs to restore muscle integrity [78]. Common ABPTs include overground locomotor training and/or bodyweight-supported treadmill training (BWSTT) that is accompanied by manual or robotic-assisted placement of the impaired limbs into normal gait patterns, passive cycling, or functional electrical stimulation (FES) that is coupled with cycling, rowing, or resistance training (RT). With intense repetitive training, ABPTs are theorized to activate and optimize sublesional spinal networks, enabling the improved performance of task-specific motor activities [79]. For example, BWSTT and other locomotor-based modalities have shown promise in promoting use-dependent neuroplasticity after mild to moderate motor-incomplete SCI and result in improved walking speed, temporal gait parameters, and lower limb muscle activation patterns [80][81][82], along with increased muscle strength and rate of torque development in some persons with incomplete SCI [78]. Structural and functional plasticity likely drives motor recovery resulting from these ABPTs and stems from the reorganization of both supraspinal and spinal cord neural circuits [83,84]. As evidence, beneficial adaptations to spinal neuronal pathways have been observed in response to ABPT by probing soleus Hoffmann reflex during walking in which homosynaptic facilitation normalized, homosynaptic depression reversed, and presynaptic inhibition of Ia afferents improved [83,85,86]. The functional recovery in persons with incomplete SCI undergoing ABPT has also been associated with findings that indicate greater descending corticospinal drive, including increased ankle dorsiflexor and knee extensor maximal motor-evoked potentials, a probe of corticospinal tract excitability [87], and improved ankle dorsiflexor and plantar flexor muscle co-activation patterns during walking [88]. Possible mechanisms underlying ABPT-mediated neuroplasticity may involve the upregulation of brain-derived neurotrophic factor and/or its receptor, tyrosine kinase B mRNA, in the spinal cord, which mediate improvements of synaptic transmission, axon regeneration, and motor neuron survival [89]. For further discussion on this topic, we refer readers to the following review [78].

Effects of ABPT and Reloading Modalities on Bone after SCI
In uninjured persons, weight-bearing exercise that produces high peak strains and/or high strain rates increases BMD [90] and prevents disuse-mediated bone loss [91]. In vari-ous models, cyclical loading has also been shown to stimulate bone formation and increase bone mass in a manner that is dependent upon the compressive strain that is applied to bone, the loading frequency (Hz), and the loading duration. In contrast, stationary/static loading does not typically alter bone parameters [92]. This knowledge has contributed to an emphasis on ABPTs to increase BMD after SCI [17] and to recommendations from healthcare providers to utilize weight-bearing activities to improve BMD after SCI [93]. However, evidence demonstrating the skeletal benefits of ABPTs is sparse in persons with SCI with several meta-analyses concluding that insufficient evidence exists to establish their effectiveness in improving BMD [94,95]. Likewise, the existing evidence is contradictory when assessing the ability of ABPTs to alter bone turnover in a manner that would be expected to improve BMD after SCI. For example, Bloomfield et al. [96] reported that nine-months of FES cycling increased serum osteocalcin (bone formation marker) >75% in persons with chronic complete SCI. Similarly, Mobarke et al. [97] reported that 12-weeks of BWSTT starting at 50% bodyweight support and progressing to full weight-bearing increased osteocalcin when compared with stretching and RT exercises, while others have reported that sixto eight-months of BWSTT suppressed urinary deoxypyridinoline/creatinine (bone resorption marker) without altering osteocalcin [50]. In comparison, Craven et al. [98] reported that those with chronic incomplete SCI displayed no change in circulating osteocalcin, CTX (bone resorption marker), or sclerostin in response to four-months of FES assisted BWSTT. Similarly, others reported that persons with incomplete to complete SCI exhibit little change in circulating bone formation markers or circulating/urinary bone resorption markers in response to FES-based cycling [30] or RT [51,99], BWSTT [100], seated [101] or standing vibration [102], or a multimodal ABPT regimen [103].
Relatively few preclinical SCI studies have evaluated the possibility that ABPTs alter bone turnover or improve BMD, likely because few preclinical modalities exist to reload the paralyzed limbs. In this regard, Qin et al. developed an implantable electrical stimulation (ES) system that elicited unilateral near-maximal contractions of the paralyzed soleus and plantaris muscles in spinal transected rats, with ES delivered 60 min/day for one-week (40 Hz at 1.5 V, (2 s on/18 s off), 1.5 V) [64]. This brief ES protocol improved plantaris muscle mass but did not alter BMD or bone microstructure when compared with SCI rats not receiving ES. However, ES suppressed circulating CTX by~50% and reduced the number of osteoclast-like TRAP-positive osteoclasts that develop in ex vivo cultures, suggesting ES suppressed osteoclastogenesis and bone resorption. ES also reversed the suppression of OPG mRNA in ex vivo osteoblast cultures, without altering RANKL mRNA, and suppressed mRNA expression of SOST, along with several other genes that negatively regulate Wnt-signaling. Despite these changes, ES did not reverse the SCI-induced suppression of osteocalcin nor did it alter the reduction in alkaline phosphatase-positive or von Kossastained colonies that develop in ex vivo osteoblast cultures that are derived from tibia or femur bone marrow, suggesting that one-week of ES did not stimulate osteoblastogenesis or bone formation. More recently, the same group reported that a four-week ES protocol lessened distal femur and proximal tibia aBMD loss and increased several trabecular and cortical bone microstructural variables at the distal femur and femoral midshaft [104]. This four-week ES protocol also increased the circulating osteocalcin, suggesting an increased whole-body bone formation. However, distal femur mineralizing surface (index of active bone formation), mineral apposition rate (index of osteoblast activity), and surface-level bone formation were similarly suppressed in SCI+ES and SCI groups vs. the controls, providing no direct evidence that ES stimulated bone formation at this site. Furthermore, ES did not prevent the suppression of osteoblastogenesis that resulted from SCI, nor did it increase osteoblast SOST mRNA expression. However, ES reduced the RANKL:OPG mRNA ratio within the hindlimbs, reduced the number of TRAP+ osteoclasts in ex vivo cultures, and produced a 50% reduction in the percentage of trabecular bone that was covered by osteoclasts when compared with the untreated SCI animals, providing evidence of an antiresorptive effect.
Beyond the studies that are described above, we are aware of only a few others assessing bone responses to loading in preclinical SCI models. Zamarioli et al. [105] evaluated two loading protocols in spinal transected rats: (1) ES of the quadriceps and triceps surae muscles [300 µs pulses, 50 Hz (5 s on/15 s off), 20-150 mA, 3 days/week, 30 min/day] and (2) a custom-built bipedal standing frame that resulted in 70% of total body mass being supported by the hindlimbs. Neither modality prevented distal femur or proximal tibia aBMD loss nor lessened the reduction in bone strength at these sites, although, the maximal strength of the lumbar vertebrae increased by~30% in response to bipedal standing. We have also reported that hindlimb cast immobilization worsened trabecular and cortical bone loss at the distal femur and proximal tibia and reduced the distal femur bone strength in a severe SCI model [39]. In this study, two-weeks of quadrupedal (q)BWSTT reversed the bone loss that resulted from cast immobilization but did not lessen bone loss that resulted directly from SCI. In a follow-up study, three-weeks of qBWSTT was also shown not to attenuate trabecular bone deterioration after SCI [59].
Based on the evidence that is presented above, it remains unknown whether ABPTs that reload the impaired limbs after SCI alter bone turnover in a manner that is sufficient to improve BMD. With the above considered, the next sections summarize the findings of clinical studies that met the following criteria: (1) enrolled adults with incomplete and/or complete SCI and (2) reported BMD at the distal femur or proximal tibia, or BMD or T-scores at other sublesional sites, both before and after an ABPT intervention regardless of other study characteristics. In doing so, we considered several important questions. First, did the study enroll persons with acute to subacute (0-12-months) SCI when bone loss is the most rapid, or during the chronic (>12-month) period when bone loss is more gradual? Secondly, did the study assess the bone changes at the relevant fracture-prone sites after SCI (i.e., distal femur or proximal tibia) or at other bone sites? Thirdly, did the study report that an ABPT attenuated bone loss or increased BMD? We considered trials that reported attenuated bone loss or increased BMD vs. the baseline or a control group to have demonstrated improvement, while those not meeting these criteria were considered to have shown no effect. For studies that combined an ABPT with a pharmacologic intervention, we only discuss the groups that were receiving ABPT alone. Case studies, case series, and cross-sectional studies that did not statistically assess the pre-post differences are discussed from a qualitative perspective.

ABPT Interventions and BMD after Acute/Subacute SCI
A visual summary of the studies examining the effects of different ABPT modalities at knee and non-knee sites in acute and subacute SCI is presented in Figure 1A,B. Case studies or case series that enrolled persons in the acute to subacute post-SCI period utilized BWSTT with manual- [50] or robotic-assistance [38], or FES rowing [106]. Of these, none reported increased BMD vs. baseline. Only FES rowing appeared to attenuate the distal femur BMD loss when compared with the expected bone loss rate during the subacute SCI phase [106], while neither BWSTT with manual- [50] or robotic-assistance [38] produced an apparent effect on BMD (Table 1). Using a cross-sectional design, Goemaere et al. [107] also reported that those with incomplete or complete SCI who initiated passive standing during the acute SCI phase displayed less aBMD deficits at the lumbar spine [0% difference (standing) vs. −7.4% (non-standing)] and femoral shaft [−21% (standing) vs. −29.2% (non-standing)] when compared with normative aBMD values that were derived from a non-SCI cohort, although, standing did not appear to attenuate aBMD deficits at the total hip or femoral neck or trochanter. imal tibia, while undergoing a six-month multimodal ABPT regimen that involved lower extremity RT, BWSTT, overground walking, vibration training, and FES cycling. Similarly, 5 of the 11 controlled trials did not detect attenuated sublesional BMD loss at any skeletal site that was examined [51,99,110,111,114]. These trials ranged from 6-weeks to 12-months in duration (training frequency three to five days/week) and used the following modalities: passive unilateral standing [110], BWSTT with standing [114], and FESbased RT [51,99,111]. Of these, only Groah et al. [51] examined BMD at the knee. The total number of studies that reported decreased BMD, no BMD change, attenuated BMD loss, or improved BMD after an ABPT modality are provided in the corresponding bars. Some studies examined both knee and non-knee sites -in these instances, the individual results for knee and non-knee sites were included in each bar. If a study compared more than one modality, the individual results for each modality were included if available. If no studies were found that fit a certain category, the category was omitted from the chart. Overall, most acute studies showed no effect or attenuated BMD loss over time. FES, functional electric stimulation; RT, resistance training; "Standing+" refers to interventions that combined standing modalities with other modalities that increase muscle activation such as FES or vibration. The total number of studies that reported decreased BMD, no BMD change, attenuated BMD loss, or improved BMD after an ABPT modality are provided in the corresponding bars. Some studies examined both knee and non-knee sites -in these instances, the individual results for knee and non-knee sites were included in each bar. If a study compared more than one modality, the individual results for each modality were included if available. If no studies were found that fit a certain category, the category was omitted from the chart. Overall, most acute studies showed no effect or attenuated BMD loss over time. FES, functional electric stimulation; RT, resistance training; "Standing+" refers to interventions that combined standing modalities with other modalities that increase muscle activation such as FES or vibration.
No trial without a control comparator [103,108] nor any controlled trial [28,31,51,99,[109][110][111][112][113][114][115] that enrolled persons in the acute/subacute period reported increased sublesional BMD vs. the baseline, irrespective of the site that was evaluated ( Table 2). There were two uncontrolled trials that reported no obvious attenuation of sublesional bone loss after ABPT. Rodgers et al. [108] reported that distal tibia trabecular vBMD was 5.4% lower vs. the baseline in a cohort of persons with subacute or chronic SCI involved in a 12-week FES RT protocol, with the subacute SCI participant displaying the greatest bone loss (−26% vs. the baseline) among all the participants. Astorino et al. [103] also reported that a cohort of persons with acute/subacute SCI or chronic SCI exhibited progressively lower aBMD at most sites that were examined, including the distal femur and proximal tibia, while undergoing a six-month multimodal ABPT regimen that involved lower extremity RT, BWSTT, overground walking, vibration training, and FES cycling. Similarly, 5 of the 11 controlled trials did not detect attenuated sublesional BMD loss at any skeletal site that was examined [51,99,110,111,114]. These trials ranged from 6-weeks to 12-months in duration (training frequency three to five days/week) and used the following modalities: passive unilateral standing [110], BWSTT with standing [114], and FES-based RT [51,99,111]. Of these, only Groah et al. [51] examined BMD at the knee.
Note: BMD determined on N = 8 <1 yr SCI and N = 5 chronic SCI. Values are avg of the total cohort and were not determined separately for acute and chronic SCI cohorts.
vs. untrained G, group; BWSTT, bodyweight-supported treadmill training; RT, resistance training; NMES, neuromuscular electrical stimulation; FES, functional electrical stimulation; F, female; M, male; C, cervical; T, thoracic; L, lumbar; AIS, American Spinal Injury Association Impairment Scale; SCI Duration, time since SCI in relation to intervention; aBMD, areal bone mineral density; vBMD, volumetric bone mineral density; min, minute; h, hour; d, day; wk, week; mo, month; yr, year; N/R, not reported; F/U, follow-up after intervention complete; Note: % change was reported in individual papers or was manually calculated from data in tables and/or figures; * indicates <0.05, ** <0.01 vs. the baseline; # indicates <0.05 vs. the initial BMD assessment after the baseline; † indicates a p-value of <0.05, † † <0.01 between the marked groups; a lack of symbols indicates no statistical differences that were reported versus the baseline or between groups.
In comparison, 6 of the 11 controlled trials reported that an ABPT attenuated BMD loss at different sublesional sites [28,31,109,112,113,115]. Lai et al. [31] reported that 3-months of FES cycling attenuated aBMD loss at the distal femur (−2.2% from the baseline) as compared to the controls (−6.7% from the baseline). Dudley-Javoroski et al. [113] reported that FES-mediated quadriceps contractions attenuated distal femur vBMD loss in the limb that was undergoing unilateral standing (25% less bone loss at 1-year and 39% less bone loss at three-years vs. the non-stimulated limbs), with a follow-up analysis confirming that FES attenuated trabecular vBMD loss in a roughly similar pattern in the anterior and posterior distal femur quadrants [116]. Shields et al. [28] also reported~50% less proximal tibia aBMD loss in the leg that underwent >1.6-years of NMES (−17% from the baseline) vs. the untrained leg of the same participant (−32% from the baseline). Additionally, three controlled trials reported attenuated BMD loss at sites other than the knee. Shields and Dudley-Javoroski [115] reported a 31% higher trabecular vBMD at the distal tibia epiphysis in the limb that received FES strengthening 5 days/week for two-to three-years vs. the untrained limb and De Bruin et al. [112] reported that BWSTT or use of standing frames lessened distal tibia trabecular vBMD loss over six-months (−0.4% from the baseline for both interventions) vs. the controls (−8% from the baseline). Lastly, Alekna et al. [109] reported that passive standing (>one hour/day, five days/week, for two-years) lessened aBMD loss vs. controls within the whole leg (−25% vs. −34% from the baseline), and pelvis (−15% vs. −21% from the baseline). Regardless, five of six controlled trials that observed attenuated BMD loss at the sites that were described above also reported no BMD differences in the treatment vs. control groups at other sites [28,31,109,112,115]. Of note, four of six trials that reported attenuated BMD loss utilized FES exercises [28,31,113,115] and two used standing exercise without FES [109,112]. However, a similar number of FES [51,99,111] and standing trials [110,114] observed no improvement in BMD at any site that was evaluated.

ABPT Interventions and BMD after Chronic SCI
A visual summary of studies examining the effects of different ABPT modalities at knee and non-knee sites in chronic SCI is presented in Figure 2A,B. Case studies or case series that enrolled participants with chronic SCI ranged from 10-weeks to 8-years in duration, instituted training frequencies of two-to five-days/week, and utilized the following modalities (Table 3): passive standing in a frame that was combined with leg or whole-body vibration [117], overground walking that was assisted by reciprocating gait orthosis [118], BWSTT alone [119] or in combination with epidural electrical stimulation [120] or nerve stimulation [121], or FES-based cycling [122,123] or rowing [124] that was delivered alone or after FES RT [106,125,126]. These reports produced inconsistent results with four case studies [117,121,123,125] and one case series [118] observing an apparent BMD increase ranging from 2-20% vs. the baseline, three case studies [121,122,124] and one case series [120] reporting no clear bone change (<2% change vs. baseline), and two case studies [117,119] and three case series [106,118,126] observing an apparent BMD reduction ranging from 2-21% vs. the baseline, depending on the site that was evaluated. Goktepe et al. [127] also reported no differences in T-scores among persons with chronic SCI who performed ≥one hour/day or <one hour/day standing or no standing at several traditional osteoporosis sites. Of these studies, only two analyzed BMD changes at the knee. Lambach et al. [106] reported distal femur trabecular vBMD was~13% lower in two persons with chronic SCI after 9-12 months of FES RT and rowing, while Coupaud et al. [121] reported inconsistent distal femur and proximal tibia total and trabecular vBMD changes (range: −2.8% to +2.5% vs. the baseline) in a person with chronic SCI who underwent BWSTT with unilateral peroneal nerve stimulation for seven-months. Within a cross-sectional design, Gibbons et al. also reported that a person with chronic SCI displayed 31% higher proximal tibia trabecular vBMD after completing eight-years of FES rowing [124] and 80-125% higher total and trabecular vBMD at the distal tibia after 10-years, [15] when compared with a historical cohort of persons with chronic SCI, although, vBMD remained 7-19% lower in this person compared with uninjured persons. Of note, three of five case studies/series that reported higher BMD vs. a baseline incorporated FES [123,125] or nerve stimulation [121], while four others that utilized FES reported no BMD change [122,124] or reduced BMD vs. the baseline [106,126].
Inconsistent results also exist in uncontrolled and controlled trials that evaluated the effects of ABPT on bone in persons with chronic SCI. In total, we identified 14 trials without a control comparator group (Table 4), which ranged from 6-weeks to 16-months in duration with a training frequency of two-to seven-days/week and assessed: passive loading in a standing frame [128], standing that was combined with whole body vibration [102], BWSTT alone [100], over ground walking in an exoskeleton [129] or reciprocating gait orthosis [130], a multimodal ABPT regimen (described above) [103], or FES-based overground walking [131], RT [108], cycling [29,30,132], or a combination of RT and cycling [133][134][135]. Three uncontrolled trials reported increased BMD [29,30,134] vs. a baseline at some sublesional sites, while 10 reported no bone change at any sublesional site evaluated [100,102,[128][129][130][131][132][133][134][135] and one enrolling persons with acute and chronic SCI observed BMD reductions at most sites assessed [103]. In addition, we identified 10 controlled trials ( Table 5) that ranged from 12-weeks to 12-months in duration, utilized a training frequency of three-to five-days/week and evaluated: seated vibration to the lower limbs [101,136], a combined protocol involving BWSTT, overground walking, and RT [97], or FES while undergoing BWSTT [98], RT [137,138], cycling [139], or RT and cycling [96,140] or rowing [141]. A total of two controlled trials reported increased BMD vs. a baseline at select bone sites [96,137] and two indicated higher BMD vs. a control group [97,140], while five reported no difference vs. a baseline or a control group at any site evaluated [98,101,136,138,139] and one acquired but did not report BMD [141]. studies/series that reported higher BMD vs. a baseline incorporated FES [123,125] or nerve stimulation [121], while four others that utilized FES reported no BMD change [122,124] or reduced BMD vs. the baseline [106,126]. Inconsistent results also exist in uncontrolled and controlled trials that evaluated the effects of ABPT on bone in persons with chronic SCI. In total, we identified 14 trials without a control comparator group (Table 4), which ranged from 6-weeks to 16-months in duration with a training frequency of two-to seven-days/week and assessed: passive loading in a standing frame [128], standing that was combined with whole body vibration [102], BWSTT alone [100], over ground walking in an exoskeleton [129] or reciprocating gait orthosis [130], a multimodal ABPT regimen (described above) [103], or FES-based overground walking [131], RT [108], cycling [29,30,132], or a combination of RT and cycling [133][134][135]. Three uncontrolled trials reported increased BMD [29,30,134] vs. a baseline at some sublesional sites, while 10 reported no bone change at any sublesional site evaluated [100,102,[128][129][130][131][132][133][134][135] and one enrolling persons with acute and chronic SCI observed BMD reductions at most sites assessed [103]. In addition, we identified 10 controlled trials ( Table 5) that ranged from 12-weeks to 12-months in duration, utilized a training frequency of three-to five-days/week and evaluated: seated vibration to the lower limbs [101,136], a combined protocol involving BWSTT, overground walking, and RT [97], or FES while undergoing BWSTT [98], RT [137,138], cycling [139], or RT and cycling [96,140] or rowing [141]. A total of two controlled trials reported increased BMD vs. a baseline at select bone sites [96,137] and two indicated higher BMD vs. a control group [97,140], while five reported no difference vs. a baseline or a control group at any site evaluated [98,101,136,138,139] and one acquired but did not report BMD [141].  Figure 1A,B. Overall, most studies on persons with chronic SCI reported that ABPT did not alter BMD with only four studies reporting improved BMD at the knee. FES, functional electric stimulation; RT, resistance training; "Walking+" refers to interventions that combined walking ABPT with FES or nerve stimulation; "Standing+" refers to interventions that combined standing modalities with other modalities that increase muscle activation such as FES or vibration. controlled and cohort studies. The data are reported as described in Figure 1A,B. Overall, most studies on persons with chronic SCI reported that ABPT did not alter BMD with only four studies reporting improved BMD at the knee. FES, functional electric stimulation; RT, resistance training; "Walking+" refers to interventions that combined walking ABPT with FES or nerve stimulation; "Standing+" refers to interventions that combined standing modalities with other modalities that increase muscle activation such as FES or vibration. Table 3. Case studies, case series, and cross-sectional studies evaluating the effects of activity-based physical therapy (ABPT) and/or loading on bone mineral density (BMD) or T-scores in adults with chronic spinal cord injury (SCI).         Note: BMD determined on N = 1 <1 yr and N = 7 chronic SCI. Values are an average of total and chronic cohorts. Acute cohort data are in Table 2. BWSTT, bodyweight-supported treadmill training; RGO, reciprocating gait orthosis; FES, functional electrical stimulation; RT, resistance training; OGW, overground walking; F, female; M, male; C, cervical; T, thoracic; L, lumbar; AIS, American Spinal Injury Association Impairment Scale; SCI Duration: time since SCI in relation to intervention reported as range, mean ± SD, or mean and (range); aBMD, areal bone mineral density; vBMD, volumetric bone mineral density; avg, average; min, minute; h, hour; d, day; wk, week; mo, month; yr, year; N/R, not reported; F/U, follow-up after intervention complete; Note: % change was reported in individual papers or was manually calculated from data in tables and/or figures; * indicates a p-value of <0.05 vs. the baseline; # indicates <0.05 vs. the initial BMD assessment after the baseline; a lack of symbols indicates no statistical differences that were reported versus the baseline or between groups. Table 5. Controlled interventional studies evaluating the effects of activity-based physical therapy (ABPT) and/or loading on bone mineral density (BMD) or T-scores in adults with chronic spinal cord injury (SCI).
Hangartner et al.    In total, five uncontrolled [13,29,30,100,103] and nine controlled trials [96,98,101,[136][137][138][139][140][141] evaluated distal femur and/or proximal tibia BMD. Most of these studies reported neither an attenuation of BMD loss nor increased BMD at areas surrounding the knee [96,98,100,101,103,136,138,139,141]. However, a small subset of these trials reported improved BMD at the knee, with each study using FES-based RT and/or cycling [29,30,134,137,140]. For example, the distal femur and/or proximal tibia aBMD increased 10-15% vs. the baseline in studies that utilized 6-18 months of FES knee extension RT (distal femur: +18%, proximal tibia: +15%) [137] or cycling (distal femur: +11%, proximal tibia: +10-13%) [29,30]. Similarly, the total and trabecular vBMD at the distal femur were improved 7% to 13% vs. the baseline, respectively, in a study that utilized three-months of FES knee extension RT followed by nine-months of FES cycling [134], suggesting that trabecular bone responds more robustly to FES than cortical bone. In this regard, Hangartner et al. reported that 12-weeks of FES knee extension RT followed by 12-to 36-weeks of FES cycling produced an estimated 3.3% less proximal tibia trabecular vBMD loss per year when compared with non-exercised controls, while no apparent cortical vBMD improvement existed [140]. A total of two trials that incorporated bodyweight-supported [97] or full weight-bearing activity without FES [103] also observed an increased aBMD at sites other than the knee. Mobarake et al. [97] reported that a 12-week BWSTT regimen increased the lumbar spine and femoral neck aBMD by +9% and +17% from the baseline, respectively, vs. a +1.2-1.3% change for the controls. Similarly, Astorino et al. [103] reported that a six-month multimodal ABPT regimen increased lumbar spine aBMD by +4.7% vs. the baseline. However, as indicated above, seven trials that assessed FES-based modalities [108,132,133,135,138,139,141] and several that incorporated standing without FES [102,128] or exoskeleton/assisted walking/BWSTT without FES [100,129,130] reported no BMD improvement, irrespective of the site that was examined.

Common Parameters to Improve BMD after SCI
Several common parameters existed among studies that reported improved proximal tibia or distal femur BMD or increased BMD at other sites. First, studies that reported attenuated BMD loss at the knee all enrolled persons with acute/subacute SCI and used FES-based modalities that were performed 3-5 days/week, 20-60 min/day, for ≥3-months [28,31,106,113] and studies reporting increased BMD at the knee all enrolled persons with chronic SCI and used FES-based regimens performed 3-5 days/week, 30+ min/day, for ≥6-months [29,30,134,137]. Second, no study that incorporated standing without FES observed improvements in the distal femur or proximal tibia BMD, although, some that utilized these regimens 3-7 days/week, 60+ min/day, for >3-months reported attenuated BMD loss [107,109,112] or increased BMD at other sites [97,103]. These observations highlight the need to target ABPTs to the specific region(s) where BMD improvements are most warranted and to use a training intensity, frequency, and duration that is sufficient to improve BMD. Regardless, it is important to reiterate that not all FES modalities or passive/active standing regimens that met these criteria produced BMD improvements. For example, Clark et al. [111] and Arija-Blazquez et al. [99] reported that FES-based RT regimens lasting 14-weeks to 5-months (performed 5-days/week for 30+ min/day) did not attenuate sublesional aBMD loss in persons with acute/subacute SCI and other FES studies that met the chronic criteria reported no BMD improvement at the distal femur or proximal tibia [96,138,139], or at other skeletal sites [108,131,135].
Several other factors that were associated with BMD improvements were also identified. First, higher total work output was associated with greater BMD gain. For example, Lambach et al. [106] reported that a higher magnitude of loading and more weekly training sessions were both associated with less trabecular vBMD loss at the proximal femur in persons that were undergoing FES rowing, and Draghici et al. [142] reported that total distance rowed and peak foot reaction force were positively associated with the preservation of the distal tibia trabecular thickness in persons that were undergoing FES rowing, when these factors and SCI duration were incorporated into stepwise regression models.
Similarly, Bloomfield et al. [96] reported that distal femur aBMD was improved (+17.8% vs. the baseline) in a subgroup of persons who achieved power outputs >18 watts (W) during FES cycling, while aBMD was not improved in those with lower power output (<12 W, BMD change +0.2%) or in non-exercising controls (−2.3% vs. the baseline) and Frotzler et al. [134] reported that FES cycling improved the total and trabecular vBMD at the distal femur with participants averaging~18 W power output. Regardless, 18 W may be difficult to sustain for many persons with SCI. As evidence, Eser et al. [143] reported that power outputs ranged from 10.8 to 13.6 W in persons with acute SCI who completed an average of 47 FES cycling bouts, with higher FES frequency producing higher power outputs. Of these participants, less than one-third achieved an average power output >18 W. Second, a continued training stimulus is required to maintain BMD if skeletal improvements are achieved. As evidence, in two studies that reported FES increased aBMD vs. the baseline, a regression of BMD to the baseline occurred after a six-month period involving no FES cycling [29] or once weekly FES [30].

Future Directions
As discussed above, only limited evidence supports the notion that ABPTs attenuate BMD loss in persons with acute/subacute SCI or increase BMD in the paralyzed limbs after chronic SCI, especially when considering the most fracture-prone sites that surround the knee. Moreover, few studies have evaluated whether ABPTs improve trabecular or cortical bone microstructure at the distal femur or proximal tibia (Table 6) and none have assessed whether the potential BMD gains or bone microstructural changes mitigate the fracture risk. Interestingly, Fang et al. [144] and Edwards et al. [101] both used FEA to predict the bone mechanical characteristics in response to 12-months of FES rowing or seated vibration, respectively, and observed no tensile or compressive strength improvements at the distal femur or proximal tibia. However, neither study observed BMD or bone microstructural changes at areas surrounding the knee in response to the respective modalities. Given these observations, we suggest future research focuses on optimizing the FES parameters to improve the distal femur and proximal tibia bone parameters, with emphasis on the following questions:

•
What are the optimal FES and modality-specific parameters (e.g., stimulation frequency, amplitude, pulse width, power output, tibio-femoral strain, and pedaling cadence) to stimulate and/or maintain bone improvements? • What is the optimal training frequency (number of sessions per day or per week) and training duration (time per bout and intervention length) to improve bone parameters? • What is the minimum training frequency and duration to maintain bone after SCI? • Is the magnitude of BMD or bone microstructural gains that result from FES sufficient to improve bone strength or bone mechanical characteristics and/or to reduce the fracture risk at the distal femur and/or proximal tibia after SCI? Clinical trials focused on optimizing FES would require a large, concerted effort over many years to discern optimal parameters, due to the relatively slow bone remodeling in humans and the number of potential parameter combinations. To lessen this burden, percutaneous ES [105,145,146] or direct nerve stimulation [64,104,147] techniques could be used to optimize the FES parameters and to identify the optimal training frequency and duration in preclinical SCI models, with only the most successful protocols advancing. Preclinical SCI models can also be used to evaluate combinatory strategies to improve BMD. In this regard, several recent preclinical and clinical studies have assessed ABPTs combined with antiresorptive drugs, such as zoledronic acid [141,144] or testosterone [59,104,148], or bone anabolic drugs such as teriparatide [101], with varying success.
Regardless of the intervention that is used, we suggest the following practices be implemented when evaluating bone changes after SCI. First, studies should quantify the distal femur and/or proximal tibia bone parameters before and after an intervention using an established imaging modality (e.g., DXA or pQCT) and a validated protocol, instead of assessing surrogate bone sites. We recognize that DXA is used to quantify whole-body fat-free mass or fat mass changes, and thus, aBMD that is assessed in the whole-body or at other traditional osteoporosis sites may be reported out of convenience. Future work should explain the relevance of BMD changes at these traditional osteoporosis sites in relation to the more fracture-prone distal femur and proximal tibia sites after SCI. Second, clinical trials should last a minimum of 6-12 months with ABPT performed ≥3-days/week for ≥30 min/day across the intervention, given the slow bone turnover and BMD improvements that are present in humans in response to loading. Third, a control group or control comparator (e.g., untrained limb) should be assessed to ensure the appropriate data interpretation. Fourth, interventions should attempt to quantify the tibio-femoral forces during training, as has been previously reported [149], and/or attempt to achieve average power outputs >18 W when utilizing FES cycling as regimens exceeding this threshold appear to have a higher likelihood to improve BMD [96]. Fifth, studies should assess the fracture risk and/or evaluate the bone tensile properties via FEA or another method, if available, to establish ABPT effectiveness because the reduction in bone strength over the first few months after SCI has been estimated to be three times greater than the aBMD loss as determined by DXA [36] and because any given percentage increase in whole-bone BMD resulting from loading has been associated with 10 times greater percentage increase in bone strength in preclinical models [90], suggesting that BMD changes do not reflect the bone mechanical properties that may influence non-traumatic fracture risk.

Conclusions
A visual summary of all the reviewed studies examining the effects of ABPT modalities at knee and non-knee sites in acute and chronic SCI is presented in Figure 3A,B. Various ABPTs promote neuromuscular benefits and use-dependent neuroplasticity in persons with SCI [78]. Despite this, no known ABPT completely prevents bone loss that develops in the lower extremities over the acute/subacute post-SCI period, regardless of the skeletal site that is evaluated. Moreover, no ABPT has shown universal success in increasing BMD at the highly fracture-prone sites surrounding the knee. However, a small subset of trials that evaluated FES modalities reported attenuated BMD loss at the distal femur and/or proximal tibia in persons with acute to subacute SCI and increased BMD in those with chronic SCI, suggesting such regimens hold promise. Additionally, a small subset of studies that utilized weight-bearing ABPTs without FES reported BMD gains at other sites but no BMD changes in areas near the knee. As such, research is needed to understand why ABPTs produce variable BMD responses after SCI and to optimize the training parameters for bone gain. Moreover, routine monitoring of the distal femur and proximal tibia BMD appears important to assess the fracture risk before engaging in ABPTs and to monitor bone changes at these highly fracture-prone sites. Figure 3. Summary of the effects of activity-based physical therapy (ABPT) on changes in bone mineral density (BMD) at the knee (distal femur and proximal tibia) or all other sublesional non-knee sites for persons with acute/sub-acute (A) or chronic (B) SCI in aggregate. Overall, half of all acute studies and most chronic studies showed no effect on BMD at the knee or non-knee sites. For data on specific ABPT modalities see Figure 1A,B (acute/subacute SCI) or Figure 2A

Data Availability Statement:
The data presented in this study are available on request from the corresponding author. The data are not publicly available due to data being property of the United States federal government.

Conflicts of Interest:
The authors declare no conflict of interest. Figure 3. Summary of the effects of activity-based physical therapy (ABPT) on changes in bone mineral density (BMD) at the knee (distal femur and proximal tibia) or all other sublesional non-knee sites for persons with acute/sub-acute (A) or chronic (B) SCI in aggregate. Overall, half of all acute studies and most chronic studies showed no effect on BMD at the knee or non-knee sites. For data on specific ABPT modalities see Figure 1A,B (acute/subacute SCI) or Figure 2A