Survival Prediction in Patients Treated Surgically for Metastases of the Appendicular Skeleton-An External Validation of 2013-SPRING Model.

1. Introduction

A major reason for cancer-associated death is the malignant tumours’ ability to disseminate to other organs [ 1 ]. With novel antitumour therapeutics, life expectancy of cancer patients has further improved, and the number of patients living with metastatic cancer, including bones, has increased [ 2 ]. Breast, prostate, renal cell and lung cancer as well as multiple myeloma frequently give metastases to bones, eventually causing pathological fractures [ 3 , 4 ].

Pain caused by bone metastases, impending fractures and pathological fractures are major limitations to patients’ quality of life [ 5 ], wherefore careful, multidisciplinary management is warranted [ 4 ]. Once a tumour has metastasised, treatment is usually palliative, i.e., focusing on prevention of disease progression, alleviation of symptoms and maintenance of mobility as well as quality of life [ 4 ]. Notably, in selected tumour entities as renal cell carcinoma, singular metastases—and sometimes even oligometastases—are nowadays considered as potentially curable and treated similar to primary tumours of bone [ 6 ].

Systemic antitumour agents, bone-targeted substances as denosumab and local radiotherapy constitute the foundation of therapy in patients with bone metastases [ 4 , 7 , 8 ]. Should pain be the major symptom in absence of fracture risk, palliative radiotherapy can prove beneficial [ 7 , 9 ]. However, the prolonged life expectancy also increases the risk for local tumour progression and thus ensuing impending or pathological fracture.

In these cases, the surgical treatment plan has to consider patients’ prognosis, general condition, rigidity and durability of fixation chosen, as well as anticipated rehabilitation time, which should not exceed life expectancy [ 10 ]. In order to estimate prognosis of bone metastasis patients with impending or pathological fractures prior to surgery, some prognostic models have been developed in the past. These include the (modified) Bauer Score for spinal lesions [ 11 , 12 ], the SORG Machine-Learning Algorithm (SORG-MLA) [ 13 ] and OPTIModel [ 14 ] for both spinal and extremity metastases, as well as the PATHFx model [ 15 ], the 2008-SPRING model [ 16 ] and its updated version, the 2013-SPRING model for extremity lesions [ 17 ]. The latter model uses readily available clinical variables in order to estimate patients’ survival probability within 3, 6 and 12 months after surgery. However, it was built and validated on patient cohorts undergoing (tumour) endoprosthetic reconstruction only, whilst patients treated with osteosynthesis were not considered.

Therefore, the aim of the current retrospective study was to externally validate the 2013-SPRING model using a single-centre cohort of bone metastasis patients treated both with (tumour) endoprostheses and osteosynthesis for impending or pathological long bone fractures. Furthermore, a separate validation for upper vs. lower limb metastases was performed.

2. Materials and Methods

All patients with bone metastases of the appendicular skeleton treated surgically at a single tertiary sarcoma centre between January 2000 and June 2018 were potentially eligible. Minimum follow-up was set at 12 months, or until patient death. Of initially 316 patients, 13 had to be excluded due to partially missing information or follow-up less than 12 months, resulting in 303 patients finally eligible. Mean patient age at time of surgery was 67.6 ± 11.1 years, and 140 were males (46.2%). One hundred and forty-one patients (46.5%) had undergone osteosynthesis, whilst the remaining one hundred and sixty-two patients (53.5%) had been treated with (tumour) endoprostheses ( Table 1 , Figure 1 ). The majority of metastases was located in the lower limbs (78.5%, Table 1 ).

All patients were followed up until last contact or visit at the local health care system, or until death. Median follow-up amounted to 6.3 (IQR: 2.3–21.8] months. At final follow-up, 276 patients had died of disease (91.1%).

2.1. The 2013-SPRING Model

The 2013-SPRING model as published by Sørensen et al. in 2018 was validated using the present patient cohort. This model predicts patient survival at 3, 6 and 12 months following surgery for bone metastases using logistic regression analyses for the respective endpoints [ 17 ]. Notably, the 2013-SPRING model, as well as its predecessor 2008-SPRING model [ 16 ], have been developed on cohorts of patients undergoing endoprosthetic reconstruction for impending or pathological fractures only, whilst those patients treated with, e.g., intramedullary nailing or compound plate osteosynthesis, were not included [ 17 ].

Variables required for calculation of the 2013-SPRING model were obtained from patients, i.e., underlying cancer type (grouped into slow, moderate and fast growing type, as proposed by the model’s developers [ 17 ]), haemoglobin levels two weeks to one day prior to surgery, impending or pathological fracture, presence of visceral metastases, presence of singular or multiple bone metastases (obtained on preoperative scans or scans performed up to three months postoperatively), Karnofski score (grouped in <70 or ≥70 points) and American Society of Anesthesiologists (ASA) score (grouped into ASA 1 and 2 vs. ASA 3 and 4). Furthermore, information on location of bone metastases and type of surgery performed was obtained.

2.2. Statistical Analysis

Means and medians are provided with corresponding standard deviations (SDs) and interquartile ranges (IQRs), respectively. Multivariate logistic regression analyses with endpoints 1) survival status at 3, 6 and 12 months were calculated using the proposed variables of the 2013-SPRING model. These models were validated on the entire cohort ( n = 303) as well as separately for (A) patients treated with endoprostheses ( n = 162) or (B) other surgical procedures ( n = 141) and (C) patients with metastases to the upper limbs ( n = 65) or (D) to the lower limbs ( n = 238).

Baseline differences between surgery groups (i.e., osteosynthesis vs. endoprosthesis) and bone metastasis location (upper vs. lower limb) were assessed with chi-squared tests for binary and categorical variables, as well as t-tests for normally distributed continuous variables. For evaluation of the model’s accuracy in prediction of patient death for the three time points, Brier score and area under the curve receiver operator characteristic (AUC ROC; with 95% confidence intervals (95% CIs]) were subsequently calculated. For the Brier score (between 0 and 1), lower scores indicate higher accuracy. For AUC ROC (also between 0 and 1), higher values signify better prediction. Furthermore, positive predictive values (PPVs) and negative predictive values (NPVs) at the respective time points were calculated for the entire cohort as well as subgroups. A p -value of <0.05 was considered statistically significant.

3. Results

The most common bone metastasis location was the femur in 221 cases (72.9%), followed by the humerus in 64 cases (21.1%; Table 2 ). Breast cancer was the most frequent histological subtype in 77 patients (25.4%), followed by lung cancer in 68 (22.4%) and renal cell carcinoma in 55 patients (18.2%). Of the 162 patients treated with endoprostheses, the majority had undergone hemiarthroplasty ( n = 100; 61.7%) of proximal hip or humerus, whereas 53 had received a tumour endoprosthesis (32.7%; Table 1 ).

3.1. Differences between Osteosynthesis Group (n = 141) and Endoprosthesis Group (n = 162)

There were some significant differences between patients treated with endoprostheses in comparison to those undergoing osteosynthesis, justifying the approach of additionally validating the 2013-SPRING model on the two subgroups separately. Differences included a higher proportion of patients with visceral metastases (55.4% vs. 43.2%; p = 0.035), a male predominance (54.9% vs. 39.5%; p = 0.012) and higher haemoglobin levels (7.5 ± 1.2 mM vs. 7.2 ± 1.0 mM; p = 0.015; Table 2 ) in the osteosynthesis group as compared with the endoprosthesis group. Furthermore, a significant difference in anatomical location of bone metastases with regards to surgical type was present, with most endoprosthetic reconstructions being performed in metastases to the femur (90.1% vs. 53.2%; p < 0.001). Moreover, a higher proportion of pathological fractures than impending fractures was treated with (tumour) endoprostheses (74.7% vs. 63.1%; p = 0.029; Table 2 ).

3.2. Differences between Upper Limb (n = 65) and Lower Limb (n = 238)

Between patients with upper limb and lower limb metastases, some differences were likewise found, wherefore the 2013-SPRING model was also validated in this subgroup. Patients with upper limb metastases were significantly more likely to undergo osteosynthesis ( n = 50; 76.9%) than those with metastases to the lower limbs ( n = 91; 38.2%; p < 0.001). However, the proportion of intramedullary nailing and (compound) osteosyntheses was comparable (upper vs. lower limb—58.0% ( n = 29) vs. 64.8% ( n = 59) intramedullary nailing; p = 0.423). Furthermore, patients with metastases of the upper limb significantly more often presented with pathological fractures ( n = 54; 83.1%) than patients with metastases to the lower limbs ( n = 156; 65.6%; p = 0.007; Table 2 ).

With regards to haemoglobin levels ( p = 0.058), presence of multiple bone metastases ( p = 0.428) or visceral metastases ( p = 0.294), Karnofski score ≥70 ( p = 0.176), ASA ( p = 0.087), primary cancer growth ( p = 0.573), patient age ( p = 0.090) or gender ( p = 0.163), no significant differences were found ( Table 2 ).

3.3. Validation of 3-, 6- and 12-Month Risk of Death

For the entire cohort ( n = 303), AUC ROC for risk of death prediction at 3, 6 and 12 months was 0.782 (95% CI: 0.729–0.843), 0.810 (95% CI: 0.763–0.858) and 0.802 (95% CI: 0.751–0.854), respectively ( Figure 2 ). Corresponding Brier score amounted to 0.170, 0.178 and 0.169 at 3, 6 and 12 months. PPVs and NPVs at the respective time points are listed in Table 3 .

3.3.1. Subgroup Analysis—Osteosynthesis Group ( n = 141) vs. Endoprosthesis Group ( n = 162)

For the osteosynthesis group only, AUC ROC was 0.821 (95% CI: 0.749–0.893) at 3 months, 0.841 (95% CI: 0.774–0.907) at 6 months and 0.829 (95% CI: 0.756–0.901) at 12 months ( Figure 3 ). Accuracy as estimated with Brier score was 0.160 at 3 months, 0.158 at 6 months and 0.184 at 12 months.

Risk of death prediction for the endoprosthesis group revealed AUC ROC of 0.782 (95% CI: 0.709–0.854) at 3 months, 0.809 (95% CI: 0.744–0.874) at 6 months and 0.794 (95% CI: 0.721–0.867) at 12 months ( Figure 3 ). Brier score amounted to 0.167, 0.181 and 0.169 at 3, 6 and 12 months, respectively.

PPVs and NPVs at the respective time points are listed in Table 3 .

3.3.2. Subgroup Analysis—Upper Limb ( n = 65) vs. Lower Limb ( n = 238)

For metastases located in the upper limbs, risk of death prediction showed AUC ROCs of 0.866 (95% CI: 0.778–0.953), 0.893 (95% CI: 0.815–0.971) and 0.781 (95% CI: 0.662–0.900), at 3, 6 and 12 months ( Figure 4 ). Brier scores at the respective time points were 0.138, 0.130 and 0.178.

For lower limb metastases, risk of death prediction revealed AUC ROCs of 0.762 (95% CI: 0.699–0.825) at 3 months, AUC ROCs of 0.792 (95% CI: 0.736–0.848) at 6 months and AUC ROCs of 0.814 (95% CI: 0.757–0.871) at 12 months ( Figure 4 ). Corresponding Brier scores were 0.175, 0.186 and 0.162 at 3, 6 and 12 months, respectively.

PPVs and NPVs at the three time points are visible in Table 3 .

4. Discussion

In the current study, the external validation of the 2013-SPRING model revealed a promising survival prediction at 3, 6 and 12 months after surgery for metastatic bone disease in an independent cohort of patients with bone metastases to the appendicular skeleton. Additionally, subgroup analysis for patients treated by endoprostheses (as in the original model) and osteosynthesis, as well as for patients with metastases to the upper limb and lower limb, revealed comparable survival prediction accuracy depending on treatment group and metastasis location.

Survival prediction in bone metastasis patients with impending or pathological fractures is necessary in order to select the most suitable treatment option. The utmost goal is quality of life-maintenance and -improvement, wherefore a stable situation with early mobilisation, full weight bearing and short in-hospital stay must be strived for [ 4 ].

However, patients may outlive complications associated with one or the other treatment option, especially in case metastatic bone is not (partially) removed during surgery as upon intramedullary stabilisation. Consequently, members of the Musculoskeletal Tumour Society (MSTS) agreed that life expectancy beyond 6 months in bone metastasis patients with impending or pathological fractures would justify the use of more durable implants [ 18 ]. Although indications for different surgical procedures (i.e., intramedullary nailing, (compound) plate osteosynthesis, hemiarthroplasty, tumour endoprosthesis) in the current study are in line with literature recommendations [ 10 ], these are mainly based on experience and retrospective analyses [ 10 ], whilst prospective studies investigating the most suitable surgical approach are yet to be performed [ 19 ].

In order to aid decision making, some models to predict survival in bone metastasis patients with impending or pathological fractures have been developed in the past [ 13 , 15 , 16 ], including the 2013-SPRING model [ 17 ]. Their validation—both internally and externally—is necessary in order to evaluate whether models tend to over- or underestimate remaining prognosis [ 20 ].

Herein, validation of the 2013-SPRING model using the entire patient cohort revealed a reliable predictive accuracy, with AUC ROCs > 0.780 at all three end points. Notably, in the original study by Sørensen et al. [ 17 ], higher AUC ROCs than in the present study were found, all exceeding 0.820. However, the authors only included patients treated with endoprostheses to develop the model. Thus, separate analyses on patients treated with endoprostheses or osteosynthesis were herein performed.

At baseline, some differences in variables necessary to construct the 2013-SPRING model were found depending on treatment group; patients receiving endoprostheses were rather female, more likely to have metastases in the femur, present with pathological fractures and lower haemoglobin levels, but were less likely to have visceral metastases.

Despite these marked differences and differing surgical philosophies, discrepancies in predictive accuracy were less pronounced. Interestingly, subgroup analysis on patients treated with osteosynthesis revealed even slightly better AUC ROCs and Brier scores at all three endpoints than the analysis on the endoprosthesis subgroup, even though the original model had been developed on patients treated with endoprostheses. The retrospective design, non-Scandinavian population and non-random assignment of patients to specific treatments (i.e., osteosynthesis vs. endoprosthesis) may all have contributed to this finding. Regardless of the subgroup analysed, any AUC ROC was >0.780. This is indicative of reliable survival prediction in bone metastasis patients, also in a cohort not originally considered upon model development.

Whilst one main goal of surgery for lower limb metastases is stabilisation enabling full weight bearing, this is of lesser significance in metastases to the upper limbs, as bones in this region are not primarily weight bearing [ 10 ]. Yet, torsional forces have to be considered, especially in metastases to the proximal humerus [ 10 , 21 ]. This may thus lead to differences in surgical techniques applied. Correspondingly, osteosyntheses were significantly more often performed in metastases to the upper limbs, whilst patients with metastases to the lower limbs were more likely to undergo endoprosthetic reconstruction. On the other hand, the proportion of (compound) plate osteosynthesis and intramedullary nailing was comparable between upper and lower limb metastases. Notably, significantly more patients with upper limb metastases already presented with pathological fractures than patients with lower limb lesions, which may be related to the fact that pain as a sign of impending fracture [ 22 ] more likely develops in weight bearing bones, leading to earlier patient referral. Nonetheless, further subgroup analysis depending on metastasis location revealed a reliable survival prediction accuracy in both lower and upper limb metastases, with the latter group even reaching slightly higher accuracy values.

Limitations of the study include the long inclusion period upon which patients were recruited, wherefore potential changes in prognosis due to refined (systemic) treatment options could not be considered. This may bias the validation results as treatment strategies have changed over the period. Moreover, as no prospective analysis had been performed, a certain impact of prognosis-adjusted treatment decisions cannot be eliminated.

Furthermore, this validation study focused on bone metastases to the appendicular skeleton only, since the 2013-SPRING model has been developed for patients with long bone metastases. In addition, we did not investigate emerging complications or additional treatments administered, as the main scope of this study was to externally validate the 2013-SPRING model regarding its risk of death prediction.

5. Conclusions

Surgical treatment of bone metastasis patients with impending or pathological fractures has to consider remaining life expectancy in order to select the most viable therapeutic approach. The external validation of the 2013-SPRING model using an independent cohort of patients likewise treated with endoprostheses or osteosynthesis revealed a reliable accuracy regarding survival prediction for both surgical strategies, as well as for upper vs. lower limb metastases.

Author Contributions

Conceptualization, S.S. and A.L.; Data curation, M.A.S. and E.M.; Formal analysis, M.A.S. and E.M.; Investigation, M.B., J.F., C.L.W., L.L., I.B. and J.S.; Methodology, M.A.S., E.M. and S.S.; Resources, M.S.S., M.M.P., I.B. and A.L.; Supervision, M.B., J.S., S.S. and A.L.; Validation, M.B., J.F., C.L.W., L.L., M.S.S. and M.M.P.; Visualization, M.A.S. and E.M.; Writing—original draft, M.A.S., E.M., M.S.S., S.S. and A.L.; Writing—review and editing, M.B., J.F., C.L.W., L.L., M.M.P., I.B. and J.S. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

The study was conducted in accordance with the Declaration of Helsinki, and approved by the Institutional Review Board of the Medical University of Graz (protocol code 31-038 ex 18/19; date of approval: 20 December 2018.

Informed Consent Statement

Patient consent was waived due to the retrospective design of the study.

Data Availability Statement

The data of this study are available upon reasonable request from the corresponding author.

Conflicts of Interest

The authors declare no conflict of interest.

Funding Statement

This research received no external funding.

Footnotes

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

1. Dillekås H., Rogers M.S., Straume O. Are 90% of deaths from cancer caused by metastases? Cancer Med. 2019;8:5574–5576. doi: 10.1002/cam4.2474.

2. Herbst R.S., Morgensztern D., Boshoff C. The biology and management of non-small cell lung cancer. Nature. 2018;553:446–454. doi: 10.1038/nature25183.

3. Coleman R.E. Clinical features of metastatic bone disease and risk of skeletal morbidity. Pt 2Clin. Cancer Res. 2006;12:6243s–6249s. doi: 10.1158/1078-0432.CCR-06-0931.

4. Coleman R.E., Croucher P.I., Padhani A.R., Clézardin P., Chow E., Fallon M., Guise T., Colangeli S., Capanna R., Costa L. Bone metastases. Nat. Rev. Dis. Primers. 2020;6:83. doi: 10.1038/s41572-020-00216-3.

5. Tahara R.K., Brewer T.M., Theriault R.L., Ueno N.T. Bone Metastasis of Breast Cancer. Adv. Exp. Med. Biol. 2019;1152:105–129. doi: 10.1007/978-3-030-20301-6_7.

6. Ratasvuori M., Wedin R., Hansen B.H., Keller J., Trovik C., Zaikova O., Bergh P., Kalen A., Laitinen M. Prognostic role of en-bloc resection and late onset of bone metastasis in patients with bone-seeking carcinomas of the kidney, breast, lung, and prostate: SSG study on 672 operated skeletal metastases. J. Surg. Oncol. 2014;110:360–365. doi: 10.1002/jso.23654.

7. Coleman R., Hadji P., Body J.-J., Santini D., Chow E., Terpos E., Oudard S., Bruland Ø., Flamen P., Kurth A., et al. Bone health in cancer: ESMO Clinical Practice Guidelines. Ann. Oncol. 2020;31:1650–1663. doi: 10.1016/j.annonc.2020.07.019.

8. Jiang L., Cui X., Ma H., Tang X. Comparison of denosumab and zoledronic acid for the treatment of solid tumors and multiple myeloma with bone metastasis: A systematic review and meta-analysis based on randomized controlled trials. J. Orthop. Surg. Res. 2021;16:400. doi: 10.1186/s13018-021-02554-8.

9. Rich S.E., Chow R., Raman S., Zeng K.L., Lutz S., Lam H., Silva M.F., Chow E. Update of the systematic review of palliative radiation therapy fractionation for bone metastases. Radiother. Oncol. 2018;126:547–557. doi: 10.1016/j.radonc.2018.01.003.

10. Willeumier J.J., van der Linden Y.M., van de Sande M.A.J., Dijkstra P.D.S. Treatment of pathological fractures of the long bones. Efort Open Rev. 2016;1:136–145. doi: 10.1302/2058-5241.1.000008.

11. Bauer H.C., Wedin R. Survival after surgery for spinal and extremity metastases. Prognostication in 241 patients. Acta Orthop. Scand. 1995;66:143–146. doi: 10.3109/17453679508995508.

12. Leithner A., Radl R., Gruber G., Hochegger M., Leithner K., Welkerling H., Rehak P., Windhager R. Predictive value of seven preoperative prognostic scoring systems for spinal metastases. Eur. Spine J. 2008;17:1488–1495. doi: 10.1007/s00586-008-0763-1.

13. Tseng T.-E., Lee C.-C., Yen H.-K., Groot O.Q., Hou C.-H., Lin S.-Y., Bongers M.E.R., Hu M.-H., Karhade A.V., Ko J.-C., et al. International Validation of the SORG Machine-learning Algorithm for Predicting the Survival of Patients with Extremity Metastases Undergoing Surgical Treatment. Clin. Orthop. Relat. Res. 2021;480:367–378. doi: 10.1097/CORR.0000000000001969.

14. Willeumier J., van der Linden Y., Van Der Wal C., Jutte P., Van Der Velden J., Smolle M., Van Der Zwaal P., Koper P., Bakri L., De Pree I., et al. An Easy-to-Use Prognostic Model for Survival Estimation for Patients with Symptomatic Long Bone Metastases. J. Bone Jt. Surg. 2018;100:196–204. doi: 10.2106/JBJS.16.01514.

15. Piccioli A., Spinelli M.S., Forsberg J.A., Wedin R., Healey J.H., Ippolito V., Daolio P.A., Ruggieri P., Maccauro G., Gasbarrini A., et al. How do we estimate survival? External validation of a tool for survival estimation in patients with metastatic bone disease—Decision analysis and comparison of three international patient populations. BMC Cancer. 2015;15:424. doi: 10.1186/s12885-015-1396-5.

16. Sorensen M.S., Gerds T.A., Hindso K., Petersen M.M. Prediction of survival after surgery due to skeletal metastases in the extremities. Bone Jt. J. 2016;98-B:271–277. doi: 10.1302/0301-620X.98B2.36107.

17. Sorensen M.S., Gerds T.A., Hindso K., Petersen M.M. External Validation and Optimization of the SPRING Model for Prediction of Survival after Surgical Treatment of Bone Metastases of the Extremities. Clin. Orthop. Relat. Res. 2018;476:1591–1599. doi: 10.1097/01.blo.0000534678.44152.ee.

18. Steensma M., Healey J.H. Trends in the surgical treatment of pathologic proximal femur fractures among Musculoskeletal Tumor Society members. Clin. Orthop. Relat. Res. 2013;471:2000–2006. doi: 10.1007/s11999-012-2724-6.

19. Wood T.J., Racano A., Yeung H. Surgical management of bone metastases: Quality of evidence and systematic review. Ann. Surg. Oncol. 2014;21:4081–4089. doi: 10.1245/s10434-014-4002-1.

20. Chen J.H., Alagappan M., Goldstein M.K., Asch S.M., Altman R.B. Decaying relevance of clinical data towards future decisions in data-driven inpatient clinical order sets. Int. J. Med. Inform. 2017;102:71–79. doi: 10.1016/j.ijmedinf.2017.03.006.

21. Al-Jahwari A., Schemitsch E.H., Wunder J.S., Ferguson P.C., Zdero R. The biomechanical effect of torsion on humeral shaft repair techniques for completed pathological fractures. J. Biomech. Eng. 2012;134:024501. doi: 10.1115/1.4005696.

22. Mirels H. Metastatic disease in long bones. A proposed scoring system for diagnosing impending pathologic fractures. Clin. Orthop. Relat. Res. 1989;1989:256–264. doi: 10.1097/00003086-198912000-00027.

Associated Data

Data Availability Statement

The data of this study are available upon reasonable request from the corresponding author.