During the first year after surgery, significant BMD changes were seen in all four ROI around the 130 mm cemented stem of the Zimmer® Segmental tumor prosthesis ending with a significant bone loss after 1 year of 8–15%. The bone loss was most pronounced (14–15%) in the 2 ROIs closest to the TM collar and lowest (8%) adjacent to the tip of the stem.
To our knowledge, there exist no previous reported longitudinal results of the periprosthetic bone remodeling after resection and reconstruction with the cemented Zimmer® Segmental tumor prosthesis. Only a few studies have investigated the periprosthetic bone remodeling after insertion of a tumor prosthesis [5, 8, 17]. As in the present study, Lan et al.  and Andersen et al.  found a further reduction in bone mineral with increased distance from the distal part of the stem towards the extension pieces, or prostheses, corresponding to the Gruen Zones 1, 2, 6 and 7 [20, 21]. The same pattern in BMD changes along the stem, as demonstrated by Lan et al. , was found in a cross-sectional study with a mean time of 31.8 months after surgery, using the contralateral leg as reference. However, the evaluation of BMD changes over time by Lan et al.  was based upon measurements in one selected ROI which limits comparison. Vennesma et al.  demonstrated that to obtain exact measurements of BMD changes after surgery, the operated side should always be reference and patients should be followed prospectively. Likewise, Kröger et al.  demonstrated that there are local differences in BMD between limbs and stated that BMD measurements years after surgery compared with contralateral values are invalid. The absolute and relative changes in BMD across all ROI within the present follow up are comparable to the remodeling around stems used in other tumor prostheses as demonstrated by Andersen et al. . Davis et al.  evaluated bone remodeling around the Kotz Modular Femur Tibia Reconstruction with a mean of 90.2 months after surgery and their results indicated that BMD reached a plateau. However, their study was cross-sectional using the contralateral limbs as reference and an interstudy comparison is therefore questionable.
The pattern in bone remodeling along the Zimmer® Segmental stem is corresponding to other findings after both cemented and uncemented primary hip arthroplasty [9, 21,22,23]. Bone remodeling and bone resorption adjacent to the proximal part of the stem is caused by distal transfer load of the prostheses due to the greater stiffness of the stem. Thus, the periprosthetic bone close to the artificial joint itself is more prone to stress shielding.
Several studies investigating primary hip arthroplasty reported a pronounced periprosthetic loss in BMD around the cemented and uncemented femur stem within the first 3 months after surgery followed by an increase or plateau after 6 month [15, 20]. The adaptive changes in bone remodeling caused by the surgical trauma to the bone after arthroplasty has been suggested to be long lasting despite increased postoperative activity [12, 24]. However, Brodner et al.  and Huang et al.  found increased BMD in the distal Gruen zones after 5 and 3 year follow up respectively and Korovessis et al.  found increased BMD at the greater and minor trochanter after 4 years follow-up. Our results indicate a progressive remodeling and loss in BMD after 1 year.
When comparing anatomic sites we found the most pronounced loss in BMD around the stems in proximal femur. Due to the low number of patients with proximal tibia tumor prostheses, we found those results not sufficient for comparison. Only few studies have illuminated changes in BMD of the distal femur following a stemmed femoral implant. Jensen et al.  found a significant increase in periprosthetic BMD during the first 6 months after surgery with the largest increase adjacent to the most proximal part of the stem. Our findings did not demonstrate the same pattern in BMD changes along the stem. However, the implants examined in the study by Jensen et al.  were not tumor-prostheses but regular stemmed revision total knee arthroplasty femoral components inserted without bone resection. In a finite element study, Van Lenthe et al.  studied bone loss and remodeling patterns of four femoral components: two primary TKAs and two stemmed revision prostheses with stem diameter of, respectively, 18 and 12 mm. Van Lenthe et al.  found the same pattern of bone resorption along the stem as in present study i.e. increased periprosthetic bone loss in the proximal, part of the stem, decreasing towards the distal part of the stem. We suggest the findings by Jensen et al.  are due to the described pre-operative immobilization of their patients followed by increased postoperative mobility. Patients in present study suffered to a large extend from pre-operative almost normal mobilization followed by prolonged post-operative immobility and sometimes chemotherapy. Nevertheless the indicated reduced loss in BMD around the distal femur stems compared to proximal femur stems may indicate the different mechanical load between sites.
Even though we used cemented fixation for all our prostheses with immediate weight bearing, the demonstrated progressive bone remodeling after 1-year could partly be explained by the well known required prolonged rehabilitation and immobilization after inplantation of tumor prostheses. This is due to prolonged surgery time and extensive loss of tissue. Furthermore, loss of bone stock in relation to chemotherapy is well described  and given the mean age in the present cohort, the well known age-related decay  in BMD will further affect the risk of progressive bone resorption after surgery.
It is well known from primary hip or knee arthroplasty that lesser stem stiffness, shorter stems and also coating may contribute to retain normal load transfer, and thus enhance bone preservation [14, 22, 23]. The various long-term follow-up results in periprosthetic BMD shows that adaptive bone remodeling after surgery also may contribute to better fixation as opposed to loosening and that it could depend on fixation method of the prostheses due to advantageous distribution and transmission of load. We speculate that the relative slow decrease in BMD until 1-year after surgery in all our ROI partly could be explained by the intended fixation of the TM collar with less load transfer to the tip of the stem and hence reduced stress shielding adjacent to the joint. However, inter study comparison in general is difficult due to differences in measurement of BMD, prostheses, methods of fixation and also patient cohort with regards to age, gender and comorbidity.
The average MSTS score was 22.3 (range: 14–30) 1 year after surgery. The patients scored highest in the walking and gait (average: 4.3) categories and lowest in function and supports (average: 3.3) categories.
The average MSTS score is slightly poorer compared to other studies evaluating tumor prostheses [10, 31]. Due to the need for prolonged rehabilitation after insertion of tumor prostheses, we suggest that the difference is partly caused by the relatively short follow up in our study compared to other studies. Also, we speculate that the MSTS score reflects that our cohort also comprised patients with MBD, which is often a group of patients in poor general health condition. Nevertheless, we find our results comparable to the 1-year evaluation by Andersen et al. .
To assess to what extend the periprosthetic changes in BMD were caused by stress shielding, immobilization or a general decrease in BMD for other causes, we performed DXA scans of both ankles. The immobilization of the operated limb is considered to be reflected by the decrease in BMD of the affected ankles. After 1-year, the decrease in BMD of the operated ankle was 9% and the non-operated ankle was close to baseline (2%). These findings indicate that the periprosthetic BMD changes during follow-up are caused by stress shielding combined with immobilization and to a lesser extend a general decrease in BMD.
We found a precision of BMD measurement of CV 2–5% which is slightly higher compared to Andersen et al.  evaluating the uncemented proximally Hydroxyapatite-Coated femur stem. This could partly be explained by the bone-cement interface in our measurements. Lan et al.  evaluated the Kotz Modular Femoral Tibial Reconstruction stems with screw fixation and found CV comparable to ours despite the fact, that they evaluated uncemented stems. However, their measures are based upon smaller ROI and since Gehrchen et al.  demonstrated that lesser ROI is associated with poorer precision, the smaller ROI size therefore could be an explanation. Nevertheless, we find our CV comparable to previous findings of cemented hip and knee arthroplasty which has proven to be adequate values to detect small adaptive bone remodeling changes [15, 22, 34].
Some limitations need to be addressed. Our sample size is relatively small and non-randomized. However, to the best of our knowledge randomized controlled trials, to evaluate different implants and methods of fixation for these patients, is not an option. Also, repeated measures can be biased by outside factors including outcome during follow-up. In addition, in case of missing values, repeated measure ANOVA, excludes all data of the participant. Furthermore, repeated measures is well suited for small sample size and despite 7 patients lost to follow up, we have only few missing data of those who completed 1-year data analysis follow-up and all available data was used when performing post-hoc students paired t-test. Nevertheless, to the best of our knowledge present study demonstrates the largest sample size in a prospectively designed study evaluating bone remodeling around a tumor prosthesis with 1-year follow-up.