Survival in Patients Operated on for Pathologic Fracture: Implications for End-of-Life Orthopedic Care
http://www.100md.com
《临床肿瘤学》
the Orthopaedic Surgery Service, Department of Surgery, Memorial Sloan-Kettering Cancer Center, New York
Weill Medical College of Cornell University, Ithaca, NY
ABSTRACT
PURPOSE: Life expectancy is routinely used as part of the decision-making process in deciding the value of surgery for the treatment of bone metastases. We sought to investigate the validity of frequently used indices in the prognostication of survival in patients with metastatic bone disease.
METHODS: The study prospectively assessed 191 patients who underwent surgery for metastatic bone disease. Diagnostic, staging, nutritional, and hematologic parameters cited to be related to life expectancy were evaluated. Preoperatively, the surgeon recorded an estimate of projected life expectancy for each patient. The time until death was recorded.
RESULTS: Kaplan-Meier survival analyses indicated that the survival estimate, primary diagnosis, use of systemic therapy, Eastern Cooperative Oncology Group (ECOG) performance status, number of bone metastases, presence of visceral metastases, and serum hemoglobin, albumin, and lymphocyte counts were significant for predicting survival (P < .004). Cox regression analysis indicated that the independently significant predictors of survival were diagnosis (P < .006), ECOG performance status (P < .04), number of bone metastases (P < .008), presence of visceral metastases (P < .03), hemoglobin count (P < .009), and survival estimate (P < .00005). Diagnosis, ECOG performance status, and visceral metastases covaried with surgeon survival estimate. Linear regression and receiver-operator characteristic assessment confirmed that clinician estimation was the most accurate predictor of survival, followed by hemoglobin count, number of visceral metastases, ECOG performance status, primary diagnosis, and number of bone metastases. Nevertheless, survival estimate was accurate in predicting actual survival in only 33 (18%) of 181 patients.
CONCLUSION: A better means of prognostication is needed. In this article, we present a sliding scale for this purpose.
INTRODUCTION
Decisions regarding potential surgery for metastatic disease require reliable data about patient survival and quality of life. This investigation evaluated factors that have been cited as important correlates to patient postoperative survival and compared them to surgeon preoperative survival estimates.
This information is sought for many reasons. It helps to set appropriate expectations for the patient, family, and medical staff. Data about cost, risk, and quality of life are often conflicting, but if these data could be weighed properly, it would help to define the most appropriate treatment for patients with metastatic bone disease.1-7 Utility analysis is an effective method to evaluate complex problems like this and can guide decision making by the individual patient, other stake holders, and society in general. It depends on the accurate prediction of survival, among other factors. Unfortunately, little data exist to inform these end-of-life decisions.8-18
The present study was designed to investigate the prognosis of patients with metastatic bone disease presenting to the orthopedic surgeon in need of skeletal surgical stabilization. Often the decision to operate on a patient with metastatic bone disease is strongly influenced by the expected survival of the patient.1,2,4-7,17,18 Previous studies focusing on specific primary cancers have been conducted and remain valuable tools in this regard.5,6,10,19-21 However, these studies do not provide a general approach that may be applied to this clinical scenario.
Much of the orthopedic literature on prognostication is devoted to the definition of clinical, radiologic, and biologic markers of disease in the nonmetastatic patient population. It is assumed that the same parameters are applicable across all stages of disease.17,18,21-26 The patients investigated in this series are, by definition, incurable, and many are in the terminal stage of their disease. We sought to investigate some well-recognized prognostic parameters to assess their value in determining the survival of patients who require treatment of metastatic disease to the bone. In addition, we sought to investigate how well these parameters stacked up against the clinical judgment of the surgeon and to translate this clinical evaluation into a reproducible format.11,17,18,27,28
METHODS
Between September 22, 1999 and March 13, 2003, all patients who were operated on for pathologic fracture or related indications were entered into a prospective quality-control database. Survival data were determined at two time points (April 2003 and May 2004). Approval for the use of clinical material was provided by the institutional review board.
Clinical parameters acquired included date of birth, sex, primary diagnosis, site of surgery, use of systemic therapy, and Eastern Cooperative Oncology Group (ECOG) performance status. Age was divided for survivorship analysis into patients older than 65 years and patients 65 years old and younger. For methodologic purposes, the primary diagnosis was grouped into five broad categories comprising lung cancer, breast cancer, prostate cancer, renal cancer, and other cancers. These categories represented the most prevalent of the diseases in this series. Site of surgery was classified into the upper extremity, the lower extremity, and the spine. Systemic therapy in the context of this study includes all forms of nonsurgical and nonradiotherapeutic adjuvants (eg, cytotoxic, hormonal, immunologic, metabolic, and so on) administered before or after surgery. The ECOG performance status29 was used as a clinical indicator of general functional condition. The ECOG performance status definitions are as follows: 0 for normal function, 1 for minimal functional impairment, 2 for impairment amounting to spending less than 50% of time in bed, 3 for impairment amounting to spending more than 50% of time in bed, and 4 for being completely bed bound. The performance status was measured at baseline as part of the standard evaluation of all patients on our service.
The number of bone metastases was based on radiologic assessment and bone scans. For statistical assessment of survivorship analysis, metastasis was categorized into patients with a single metastatic focus and patients who had multiple bone metastases. The presence of visceral metastases was based on routine computed tomography scans of the chest, abdomen, and pelvis. Nodal involvement was categorized for methodologic purposes into presence or absence of adenopathy and was assessed by physical examination and imaging studies of thoracic, abdominal, and pelvic cavities. Nodal involvement alone, independent of visceral involvement, was not classified as visceral metastases per se.
Blood investigations were based on the routine preoperative assessment for surgery. Laboratory data acquired included serum albumin and calcium, hemoglobin, and lymphocyte counts.10,21-27 For purposes of survivorship analysis, the following data were divided into two groups: serum albumin, 3.5 g/dL v greater than 3.5 g/dL; serum calcium, 9 mg/dL v greater than 9 mg/dL; hemoglobin, 10 g/dL v greater than 10 g/dL; and lymphocyte count, 500 cells/μL v greater than 500 cells/μL.
The authors' service represents a specialized orthopaedic oncology unit in a tertiary referral cancer center. Most patients were treated by the senior author (P.J.B.) who has close to 30 years of experience in the field. All official consults are maintained in a department-owned database. This excludes informal consults for patients who nevertheless may have had surgery and were included in this study. At the time of first consult, all patients deemed suitable for surgical stabilization were reviewed by the surgeon who made an assessment of the expected survival of the patient in months. This was recorded, and the patient was subsequently observed. By definition, the prediction was accurate if the actual survival was within 20% of the surgeon estimate of expected survival.
Data on the occurrence of death was gleaned from an institution-owned patient management database (Disease Management System version 5.2, 1996; Memorial Sloan-Kettering Cancer Center, New York, NY). Data was maintained in hard copy and electronically in a spreadsheet in Microsoft Excel version 10 for Windows NT (Microsoft, Redmond, WA). Unless otherwise stated, clinical parameters are presented as the mean ± standard deviation, and survival in months is presented as the median with 95% CIs.
Statistical analysis was performed using SPSS version 11.5 for Windows NT (SPSS Inc, Chicago, IL). Univariate analysis using the Kaplan-Meier method with log-rank assessment was performed to assess prognostic significance of individual risk factors (Figs 1 to 6). Cox regression multivariate analysis of factors found statistically significant by univariate analysis was used to assess for the independent prognostic value of risk factors. Comparisons of variables defined by means and standard deviations alone were performed with the Student's two-tailed t test. Linear regression analysis and receiver-operator characteristic (ROC) curve assessment of independently prognostic variables were used to show the relative value of these factors in predicting survival (Fig 7). For the parameter of primary diagnosis, the hazard ratio was derived from survival analysis and used as the independent variable in linear regression and ROC assessment. Statistical significance was defined as P < .05.
Data from the parameters found to be statistically significant by multivariate analysis were consolidated in a sliding scale based on median survival and 95% CI (Fig 8). To use the sliding scale, the relevant parameters were placed on the line representing the median and 95% CI in order of chronologic clinical assessment (ie, diagnosis, ECOG performance status, bone metastasis, visceral metastasis, and hemoglobin). When there was disparity between factors (eg, a patient with low hemoglobin but single metastasis), the relative position of the other factors and their statistical significance as well as the clinician estimate influenced placement.
RESULTS
Between September 22, 1999 and March 13, 2003, 191 patients were entered onto the study. All patients underwent surgical procedures pertaining to metastatic bone disease. The data were periodically updated during the 4 years of the study. The time of last accrual of data was May 19, 2004. At the time of last accrual, 50 patients remained alive, and there had been 141 deaths. Survival estimates had been provided for 181 patients. Kaplan-Meier and Cox regression analyses were performed on all 191 patients. Overall, median survival in the cohort was 8 months (95% CI, 5.4 to 10.6 months).
There were 86 female patients (45%) and 105 male patients (55%) in this series. Age was evenly distributed between the two groups (study population: mean age, 61.8 ± 13.6 years; mean for males, 61.6 ± 13.5 years; mean for females, 62.0 ± 13.7 years). The most prevalent diagnostic groups included 39 patients (20%) with lung cancer, 37 patients (19%) with breast cancer, 31 patients (16%) with renal cancer, 15 patients (8%) with prostate cancer, and 69 patients (36%) with other cancers (Table 1).
Within the stipulated time, there had been about 320 official fracture-related consults on the service. Therefore, the operative rate was 60% (191 of 320 consults). Surgery was performed on the spine in 34 patients (18%), on the upper extremity in 52 patients (27%), and on the lower extremity in 105 patients (55%). The main indications for surgery were fractures or impending fractures of bones of the extremity afflicted by metastatic disease. Impending fractures in the extremities were assessed, as per published guidelines, to be at high risk for fracture and prophylactically treated if they were osteolytic more than osteoblastic, were in the peritrochanteric area of the femur rather than in non–weight-bearing bones, involved more than half of the circumference of a bone, and were associated with functional pain.9 Joint replacement surgery was performed for fractures in the hip, knee, shoulder, and elbow. Fracture stabilization using plates, screws, and rods supplemented with cement was performed in long bones of the extremity. Spinal surgical decompression and stabilization were performed for neurologic compromise or potential neurologic compromise in the spine, especially in radiation-resistant tumors. Amputations were performed in patients with otherwise unmanageable limbs who were unsuitable for extensive surgery. Accordingly, joint replacement surgery was performed in 96 patients (50%), extremity fracture stabilization was performed in 59 patients (31%), spinal surgical decompression and stabilization were performed in 34 patients (18%), and amputations were performed in two patients (1%).
Twenty patients required reoperation. In general, this was a result of disease progression, which compromised the previous construct. Two patients required surgery for infection. Four patients had surgery for disease progression in the form of excision of soft tissue masses, which did not compromise previous constructs. Fracture stabilization failure was the most common indication for reoperation (seven patients). Other indications included five joint replacement failures and two spinal instrumentation failures.
During the course of the study, 85 patients (45%) did not receive systemic therapy, and 106 patients (55%) did receive systemic therapy. ECOG performance status was favorable (0, 1, or 2) in 93 patients (49%) and unfavorable (3 and 4) in 93 patients (49%). ECOG performance status was not available in five patients. Bone metastasis was single in 55 patients (29%) and multiple in 136 patients (71%). Nonskeletal visceral metastases were present in 113 patients (59%) and absent in 78 patients (41%). Nodal involvement was noted in 34 patients (18%) and absent in 157 patients (82%). The mean albumin level in the cohort was 3.9 ± 0.6 g/dL. The mean serum calcium measured 9.2 ± 1.0 mg/dL. The mean hemoglobin level was 11.5 ± 1.9 g/dL, and the mean absolute lymphocyte count was 1,500 ± 5,100 cells/μL.
Of the clinical parameters, only primary diagnosis, use of systemic therapy, and ECOG performance status were statistically significant predictors of survival by Kaplan-Meier analysis. As defined, age older than or 65 years was not found to be statistically significant for predicting survival (P < .8). Sex (P < .4) and site of surgery (P < .8) were similarly unremarkable for predicting survival.
The primary histologic type was significant in predicting survival (P < .008). Lung cancer patients fared the worst, with a median survival (Table 2) from the time of orthopedic consult of 4 months (95% CI, 2.2 to 5.7 months). Renal cell carcinoma patients fared the best, with a median survival time of 20 months (95% CI, 15.2 to 24.8 months). The median survivals of all other patients were clustered between the values for these two histologic types (Fig 1).
The use of systemic therapy (Fig 2) was a significant predictor of survival (P < .0004). Systemic therapy use was associated with a poorer prognosis, with a median survival time of 5 months (95% CI, 3.3 to 6.7 months). Paradoxically, patients who did not receive systemic therapy had better prognoses (median survival time, 16 months; 95% CI, 5.7 to 26.3 months).
ECOG performance status (Fig 3) was significantly predictive of survival (P < .0001). Median survival time for patients with an ECOG performance status of 0, 1, or 2 was 14 months (95% CI, 8.1 to 19.9 months), which contrasted strongly with patients with an ECOG performance status of 3 or 4 (median survival time, 5 months; 95% CI, 2.8 to 7.2 months).
Both the presence of visceral metastases and the number of bones involved by metastatic disease were significant predictors of survival. Patients with single bony metastasis (Fig 4) did significantly better than patients with multiple bony metastases (P < .00001). Median survival time was 24.7 months (95% CI, 17.5 to 32 months) for patients with single bony metastasis compared with 6 months (95% CI, 3.6 to 8.4 months) for patients with multiple bony metastases. Nodal involvement was not a significant predictor of survival (P < .08). Patients with visceral metastases (Fig 5) did significantly worse than patients without visceral metastases (P < .00001), with median survival times of 6 months (95% CI, 4 to 8 months) versus 24.7 months (95% CI, 15.2 to 34 months), respectively.
Several of the blood test results correlated with survival. Hemoglobin (Fig 6A) was a significant predictor of survival (P < .0001). Patients with a low hemoglobin level had a median survival time of 3 months (95% CI, 1.3 to 4.7 months), whereas patients with higher levels survived longer (median survival time, 10.3 months; 95% CI, 5.5 to 15.2 months). Albumin level (Fig 6B) had a significant prognostic value (P < .004). Patients with low albumin had a median survival time of 5.6 months (95% CI, 2.6 to 8.6 months), whereas patients with normal albumin fared better (median survival time, 10.3 months; 95% CI, 5 to 15.7 months). Absolute lymphocyte counts (Fig 6C) were significantly predictive of survival (P < .0003). The median survival time in patients with low lymphocyte counts was 3.2 months (95% CI, 2.3 to 4.1 months) compared with 10.3 months (95% CI, 6.4 to 14.3 months) in patients with higher counts. Serum calcium was not a significant predictor of survival (P < .8).
The parameters found to be significant by Kaplan-Meier univariate analysis were assessed for independent prognostic value using Cox regression multivariate analysis. The results of the analysis for the parameters of primary diagnosis, systemic therapy, ECOG performance status, bone metastases, visceral metastases, hemoglobin level, serum albumin level, and absolute lymphocyte count are listed in Table 2. In addition, estimated survival was included in a second multivariate analysis.
The six parameters found to be independently significant in predicting survival were primary diagnosis (P < .006), ECOG performance status (P < .04), number of bone metastases (P < .008), presence of visceral metastases (P < .03), hemoglobin level (P < .009), and survival estimate (P < .00005). Survival estimate was found to be a covariate with diagnosis, ECOG performance status, and visceral metastases. The trend 1 year later was similar, except that multiplicity of bone metastases became less significant (P < .05) and presence of visceral metastases became much more significant (P < .00003).
These six parameters were assessed using linear regression and ROC curve assessment (Fig 7). Linear regression analysis (graphs not shown) revealed a squared correlation value (R2) that was highest for estimated survival (R2 = 0.33), followed by hemoglobin (R2 = 0.15), visceral metastases (R2 = 0.11), ECOG performance status (R2 = 0.06), primary diagnosis (R2 = 0.05), and bone metastases (R2 = 0.05). Linear regression analysis (Fig 7A) of estimated survival within 20% of the actual estimate reduced the R2 value to 0.30, suggesting that the clinician's estimate could not be further optimized by simply increasing its tolerance. ROC curve assessment (Fig 7B) of survival as the state variable showed that the area under the curve and, thus, prognostic value were greatest for estimated survival, followed by visceral metastases, hemoglobin level, bone metastases, diagnosis, and ECOG performance status.
At final analysis, the survival estimate with a 20% margin of error was accurate in predicting actual survival in 33 (18%) of 181 patients. Of the remaining 148 patients, 78 (43%) were underestimated and 70 (39%) were overestimated in terms of duration of expected survival. There did not seem to be a systematic error in over- or underestimating survival. From a practical standpoint, it is often necessary to give a minimum survival estimate (ie, the minimum duration that the patient is at least likely to survive). Analysis with a 20% margin of error showed that the survival estimate was able to predict the minimum survival of 111 (61%) of 181 patients. There was no tendency towards over- or underestimation in men versus women or in patients older versus younger than 65 years. When corrected for diagnosis, the numbers in each comparison group were considerably reduced. In the major diagnostic groups, however, there was no significant trend towards over- or underestimation with respect to sex or age.
When the patients were categorized into groups of survivors, namely short (survival for 3 months), intermediate (survival > 3 and 9 months), and long (> 9 months), the data indicates that, although the clinician estimate is the most accurate prognostic parameter, it is apparently miscalibrated. Defined in this manner, 34 (64%) of 53 short survivors, 22 (53%) of 41 intermediate survivors, and 56 (64%) of 87 long survivors were categorized accurately (Fig 7A). In the short survival group, actual survival (1.4 ± 1.1 months) was significantly lower (P < .000001) than the surgeon estimate (4.5 ± 3.9 months). In the intermediate survival group, actual survival (5.9 ± 1.7 months) was also significantly lower than the surgeon estimate, which was 8.1 ± 5.7 months (P < .02). In the long survival group, actual survival (24.9 ± 10.4 months) was significantly longer than the surgeon estimate, which was 14.8 ± 9.3 months (P < .000001). Hence, short and intermediate survivors were overestimated and long survivors were underestimated in terms of expected duration of survival.
By considering all significant factors listed in Table 2, a sliding scale may be constructed (Fig 8). To test this scale, five subsequent patients not in this study were gleaned from a department-owned database, and their parameters were compared with the chart (Table 3). Figure 8B illustrates two patients from Table 3 mapped onto the chart as bands. Patient 1, who had a diagnosis in the other category and an ECOG performance status of 3, had visceral metastases, multiple bony metastases, and a hemoglobin level of more than 10 g/dL. The intersection of all these values was ultimately dictated by the upper limit of the range representing ECOG performance status and the lower limit of the range representing the hemoglobin level, yielding an estimated survival of 6 to 7 months. The actual survival was 7 months. Patient 3, who had a diagnosis in the other category and an ECOG performance status of 4, had visceral metastases, multiple bony metastases, and a hemoglobin level of less than 10 g/dL. In contrast, the intersection of all these values corresponded to the upper limit of the range representing the hemoglobin level and the lower limit of the range representing the presence of visceral metastases. This yielded a survival estimate of 4 months, which corresponded to the actual survival. As shown from these results, patients with poor prognostic factors are reasonably predicted to have poor survival. Patients who die prematurely from nononcologic underlying medical problems and patients with discordant prognostic factors are less accurately predicted, although this difference is small (2 months in patient 4 in our example). In patients with discordant factors, the clinician estimate is used as an additional parameter to guide placement on the scale.
DISCUSSION
This article attempts to answer a difficult question. What is the expected survival in patients presenting with metastatic bone disease? In their early description of treatment for this condition, Beals et al8 stated that the indication for surgery was that these "fractures were predictable." The idea that there should be a reasonable life expectancy before considering surgery is a relatively recent suggestion.1,2,4-7 In addition, being able to predict the expected survival of a patient has fundamental implications on the kind of reconstruction used. Patients with a prolonged life expectancy should receive an appropriately durable reconstruction, as opposed to the individual with a short life expectancy for whom an expedient method may be preferable.3,10
The present study has one major limitation; all patients considered for surgery had been physician selected to be in an ill-defined group with an expected survival reasonable enough to benefit from surgery. Also, issues of comorbidity are undoubtedly important but were not considered in our study. Recent work suggests that these issues should receive closer scrutiny.30 Hence, this data may not be applicable to patients who are not deemed reasonable surgical candidates or who refused surgery. Terminally ill patients with less than a 1-month prognosis were not operated on because they had a prohibitive operative risk. The main value of this analysis is in predicting survival in all other patients who were at least surgical candidates at the outset, which is the more common clinical scenario for the orthopedic surgeon.11,27,28
The main parameters assessed were gleaned from previous studies that have looked at prognostic factors in a number of conditions.10,19-22,25,31-33 The clinical parameters of age and sex were not expected to be of prognostic significance because these two parameters were evenly distributed among the good and bad performers. Perhaps surprisingly, the site of surgery was not predictive of survival.3 This has interesting implications. Spine surgery, for example, has been regarded as high-risk surgery that should be justified by an adequate estimated survival.7 Our findings suggest that this concept is misleading. The other interpretation of this data is that the successful surgical treatment of the patient's metastatic disease may improve survival to the point that the actual region involved becomes immaterial.34
Primary site was found to be a significant predictor of survival. The four most prevalent diagnostic groups (lung, breast, renal cell, and prostate cancer) are representative of most series on the subject.9,21 Patients with lung cancer fared the worst, and patients with renal cancer fared the best, as shown in Figure 1 and Table 2. Diagnosis was found to be independently predictive of survival on multivariate analysis.17,18
Systemic therapy use was found to be a poor prognostic indicator for survival. This should not be taken to mean that systemic therapy is contraindicated in this condition. Because of the complex combinations of different forms of systemic therapy in patients with prostate, breast, and lung cancer, it is difficult to interpret the meaning of this result. In this series, systemic therapy use was unevenly distributed between the primary diagnostic categories. Of the four common diagnoses, systemic therapy was predominantly used in breast and prostate patients. In lung cancer patients and all other patients (the patients with the worst prognoses), systemic therapy was used in approximately half of the patients. The overwhelming majority of renal cancer patients (the patients with the best survival) did not receive systemic therapy. This selection bias accounted for the poor prognosis associated with systemic therapy.
The ECOG performance status was found to be a significant prognostic factor. This has similarly been found to be a useful prognostic indicator in other studies. Furthermore, it's simplicity of use and reproducibility makes it a valuable index for prognostication.29,35-37
Bone metastases were found to be independently predictive of survival.4 The concept of polyostotic metastases being poorly prognostic has important implications. At least part of the morbidity and mortality that result from pathologic fractures is the predisposition to infirmity and immobility,34 which are readily amenable to improvement by surgical intervention. This would be a strong justification for surgical intervention in these patients as a means of prolonging life and not just palliation alone.
Hematologic markers have been used in various studies as prognostic indicators.21-27 Surprisingly, serum calcium was not found to be a significant predictor of survival both by univariate Kaplan-Meier and t test analyses. Hypercalcemia is a problem that is actively pursued and treated in this institution; hence, it may not have become a clinically apparent problem. This is supported by the absence of hypercalcemia seen in this series; the levels in survivors (9.27 ± 0.86 mg/dL) were not significantly different from levels in nonsurvivors (9.23 ± 1.03 mg/dL). Historically, this factor may have been important, but modern oncologic practice including bisphosphonate treatment may have neutralized this prognostic factor. Serum albumin, lymphocyte counts, and hemoglobin level were all found to be significant predictors of survival. However, only hemoglobin level was found to be an independent predictor of survival.17,18,22-25,27
Diagnosis, ECOG performance status, number of bone metastases, presence of visceral metastases, hemoglobin level, and survival estimate were independent predictors of survival.4-6,10,17-19,27,36,37 The trend 1 year later was similar, except that multiplicity of bone metastases became less significant (P < .05) and presence of visceral metastases became much more significant (P < .00003); this is likely a reflection of disease progression and increased tumor load. In general, it was found that patients with an unfavorable primary diagnosis, poor ECOG performance status, multiple bone and visceral metastases, and low hemoglobin level survived half as long as patients with favorable parameters (Fig 8). Linear regression (Fig 7A) and ROC curve assessment (Fig 7B) suggest that the predictive values of these parameters for survival were low, accounting for between 5% and 15% of the variance in the data (R2 = 0.05 to 0.15). This weak correlation implies that these parameters are not as clinically useful as commonly assumed. The attending surgeon's prediction (R2 = 0.33) was clearly superior to any of these parameters; this may be attributed to their extensive experience in the field.11,27,28 Conversely, by considering hemoglobin level (R2 = 0.15), visceral metastases (R2 = 0.11), ECOG performance status (R2 = 0.06), primary diagnosis (R2 = 0.05), and bone metastases (R2 = 0.05), 42% (the sum of R2) predictability of survival is possible from an objective standpoint. This may be superior to that offered by experience alone.
Interestingly, within specific survival groups, the numbers of short and intermediate survivors were overestimated and the number of long survivors was underestimated in terms of duration of survival. This provides some insight into conflicting literature that, for the most part, shows that survival estimates tend to be overoptimistic but has occasionally reported the reverse. Furthermore, it has been shown that, when clinical prediction of survival increased, variability in actual survival also increased.11 The accuracy of survival estimation for the surgical patients in this series illustrates this point well.
One possible explanation in this study for why diagnosis, ECOG performance status, and visceral metastasis were all significant predictors rendered insignificant by multivariate analysis is that these parameters covaried with the surgeon's estimate or that the surgeon already considered all these factors at some level in deriving a conclusion. Yet the surgeon's estimate alone was inaccurate as a predictor of survival. Perhaps the determinant is an ambiguous combination of factors that has, to date, not been determined. It may be that the patient's cognitive state, demeanor, drive, general state, or other factors together with the more traditional parameters should all be considered in the prognostication of these patients.11,18,27,28,38,39
The suggestion that the clinician estimate is accurate but miscalibrated has been described.11,18 This may be capitalized on by getting the treating clinician to suggest a prognostic category (eg, short, intermediate, or long survival) and support this estimate using the data presented in Table 2 and Figure 8. It would prove to be a more objective undertaking than suggesting an actual number, which is rarely helpful. Such a model of assessment would be a useful adjunct to the graphical system proposed. Graphical methods of prognostication are admittedly cumbersome.4,9,11,27,28,35,40,41 The proposed sliding scale system has the benefit of allowing for the influence of the clinician estimate, which, as we and others have described, is the most accurate tool in prognostication.11,17,18,28 We believe that the clinician estimate should be considered as one of a few important criteria rather than as a unique criterion in choosing therapeutic interventions.38,39 To that end, a future study will be conducted to validate this prognostic model.
In summary, this study shows that the only independent predictors of survival in the patient with bone metastases are diagnosis, ECOG performance status, number of bone metastases, presence of visceral metastases, and hemoglobin level. These parameters distinguish, with an accuracy of 5% to 15%, the ability of patients to survive less than 6 months when these parameters are unfavorable and more than 12 months when they are favorable. An assessment by a senior surgeon independent of these factors is far more accurate at 33%. This is a wake-up call to the community at large. Justification of surgery on the basis of survival prognostication is dangerously inaccurate. There is a need to create a more accurate prognostic index for patients undergoing orthopedic surgery for bone metastases.
Authors' Disclosures of Potential Conflicts of Interest
Although all authors completed the disclosure declaration, the following author or immediate family members indicated a financial interest. No conflict exists for drugs or devices used in a study if they are not being evaluated as part of the investigation. For a detailed description of the disclosure categories, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration and the Disclosures of Potential Conflicts of Interest section in Information for Contributors.
NOTES
Supported by grants from the Biomet Oncology Fellowship and the Pearlman Limb Preservation Fund.
Presented in part at the 12th International Symposium of Limb Salvage, Rio de Janeiro, Brazil, September 15-16, 2003.
This work is original and solely owned by the authors and their institution.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Harrington KD: Introduction, in: Orthopaedic Management of Metastatic Bone Disease (ed 1). St Louis, MO, Mosby, 1988, pp 1-14
Coleman RE, Rubens RD: Bone metastases, in: Abeloff MD, Armitage JO, Lichter AS, et al (eds): Clinical Oncology (ed 2). New York, NY, Churchill-Livingstone, 2004, pp 836-871
Bohm P, Huber J: The surgical treatment of bony metastases of the spine and limbs. J Bone Joint Surg Br 84:521-529, 2002
Tokuhashi Y, Matsuzaki H, Toriyama S, et al: Scoring system for the preoperative evaluation of metastatic spine tumor prognosis. Spine 15:1110-1113, 1990
James JJ, Evans AJ, Pinder SE, et al: Bone metastases from breast carcinoma: Histopathological-radiological correlations and prognostic features. Br J Cancer 89:660-665, 2003
Pietropaoli MP, Damron TA, Vermont AI: Bone metastases from squamous cell carcinoma of the head and neck. J Surg Oncol 75:136-141, 2000
Cohen DB, Riley LH: Management of metastatic carcinoma to the spine: Surgical treatment, in: Sim FH, Menendez LR (eds): Orthopedic Knowledge Update: Musculoskeletal Tumors (ed 1). Rosemont, IL, American Academy of Orthopedic Surgeons, 2002, pp 331-348
Beals RK, Lawton GD, Snell WE: Prophylactic internal fixation of the femur in metastatic breast cancer. Cancer 28:1350-1354, 1971
Mirels H: Metastatic disease in long bones: A proposed scoring system for diagnosing impending pathologic fractures. Clin Orthop Relat Res 249:256-264, 1989
Wedin R: Surgical treatment for pathologic fracture. Acta Orthop Scand Suppl 72:1-29, 2001
Glare P, Virik K, Jones M, et al: A systematic review of physicians' survival predictions in terminally ill cancer patients. BMJ 327:195, 2003
Fidler M: Prophylactic internal fixation of secondary neoplastic deposits in long bones. Br Med J 1:341-343, 1973
The Breast Specialty Group of the British Association of Surgical Oncology: British Association of Surgical Oncology Guidelines: The management of metastatic bone disease in the United Kingdom. Eur J Surg Oncol 25:3-23, 1999
Zhou Z, Redaelli A, Johnell O, et al: A retrospective analysis of health care costs for bone fractures in women with early-stage breast carcinoma. Cancer 100:507-517, 2004
Hansen BH, Keller J, Laitinen M, et al: The Scandinavian Sarcoma Group Skeletal Metastasis Register: Survival after surgery for bone metastases in the pelvis and extremities. Acta Orthop Scand Suppl 75:11-15, 2004
Miner TJ, Jaques DP, Shriver CD: A prospective evaluation of patients undergoing surgery for the palliation of an advanced malignancy. Ann Surg Oncol 9:696-703, 2002
Vigano A, Bruera E, Jhangri GS, et al: Clinical survival predictors in patients with advanced cancer. Arch Intern Med 160:861-868, 2000
Vigano A, Dorgan M, Buckingham J, et al: Survival prediction in terminal cancer patients: A systematic review of the medical literature. Palliat Med 14:363-374, 2000
Yamashita K, Koyama H, Inaji H: Prognostic significance of bone metastasis from breast cancer. Clin Orthop Relat Res 312:89-94, 1995
Rana A, Chisholm GD, Khan M, et al: Patterns of bone metastasis and their prognostic significance in patients with carcinoma of the prostate. Br J Urol 72:933-936, 1993
Coleman RE: Skeletal complications of malignancy. Cancer 80:1588-1594, 1997
Conlan DP: Value of lymphocyte counts as a prognostic index of survival following femoral neck fractures. Injury 20:352-354, 1989
Carson JL, Terrin ML, Barton FB, et al: A pilot randomized trial comparing symptomatic vs. hemoglobin-level-driven red blood cell transfusions following hip fracture. Transfusion 38:522-529, 1998
Gruson KI, Aharonoff GB, Egol KA, et al: The relationship between admission hemoglobin level and outcome after hip fracture. J Orthop Trauma 16:39-44, 2002
Koval KJ, Maurer SG, Su ET, et al: The effects of nutritional status on outcome after hip fracture. J Orthop Trauma 13:164-169, 1999
Walls J, Bundred N, Howell A: Hypercalcemia and bone resorption in malignancy. Clin Orthop Relat Res 312:51-63, 1995
Muers MF, Shevlin P, Brown J: Prognosis in lung cancer: Physicians' opinions compared with outcome and a predictive model. Thorax 51:894-902, 1996
Vigano A, Dorgan M, Bruera E, et al: The relative accuracy of the clinical estimation of the duration of life for patients with end of life cancer. Cancer 86:170-176, 1999
Oken MM, Creech RH, Tormey DC, et al: Toxicity and response criteria of the Eastern Cooperative Oncology Group. Am J Clin Oncol 5:649-655, 1982
Piccirillo JF, Tierney RM, Costas I, et al: Prognostic importance of comorbidity in a hospital-based cancer registry. JAMA 291:2441-2447, 2004
Jensen JS, Tondevold E: Mortality after hip fractures. Acta Orthop Scand 50:161-167, 1979
Kyo T, Takaoka K, Ono K: Femoral neck fracture: Factors related to ambulation and prognosis. Clin Orthop Relat Res 292:215-222, 1993
Townsend PW, Rosenthal HG, Smalley SR, et al: Impact of postoperative radiation therapy and other perioperative factors on outcome after orthopedic stabilization of impending or pathologic fractures due to metastatic disease. J Clin Oncol 12:2345-2350, 1994
Nozue M, Oshiro Y, Kurata M, et al: Treatment and prognosis in colorectal cancer patients with bone metastasis. Oncol Rep 9:109-112, 2002
Blay JY, Lasset C, Carrie C, et al: Multivariate analysis of prognostic factors in patients with non HIV-related primary cerebral lymphoma: A proposal for a prognostic scoring. Br J Cancer 67:1136-1141, 1993
Masutani M, Tsujino I, Fujie T, et al: Moderate dose-intensive chemotherapy for patients with non-small cell lung cancer: Randomized trial, can it improve survival of patients with good performance status? Oncol Rep 6:1045-1050, 1999
Mani S, Todd MB, Katz K, et al: Prognostic factors for survival in patients with metastatic renal cancer treated with biological response modifiers. J Urol 154:35-40, 1995
Pirovano M, Maltoni M, Nanni O, et al: A new palliative prognostic score: A first step for the staging of terminally ill cancer patients—Italian Multicenter and Study Group on Palliative Care. J Pain Symptom Manage 17:231-239, 1999
Maltoni M, Nanni O, Pirovano M, et al: Successful validation of the palliative prognostic score in terminally ill cancer patients: Italian Multicenter Study Group on Palliative Care. J Pain Symptom Manage 17:240-247, 1999
Culine S, Kramar A, Saghatchian M, et al: Development and validation of a prognostic model to predict the length of survival in patients with carcinomas of an unknown primary site. J Clin Oncol 20:4679-4683, 2002
Hess KR, Abbruzzese MC, Lenzi R, et al: Classification and regression tree analysis of 1000 consecutive patients with unknown primary carcinoma. Clin Cancer Res 5:3403-3410, 1999(Saminathan S. Nathan, Joh)
Weill Medical College of Cornell University, Ithaca, NY
ABSTRACT
PURPOSE: Life expectancy is routinely used as part of the decision-making process in deciding the value of surgery for the treatment of bone metastases. We sought to investigate the validity of frequently used indices in the prognostication of survival in patients with metastatic bone disease.
METHODS: The study prospectively assessed 191 patients who underwent surgery for metastatic bone disease. Diagnostic, staging, nutritional, and hematologic parameters cited to be related to life expectancy were evaluated. Preoperatively, the surgeon recorded an estimate of projected life expectancy for each patient. The time until death was recorded.
RESULTS: Kaplan-Meier survival analyses indicated that the survival estimate, primary diagnosis, use of systemic therapy, Eastern Cooperative Oncology Group (ECOG) performance status, number of bone metastases, presence of visceral metastases, and serum hemoglobin, albumin, and lymphocyte counts were significant for predicting survival (P < .004). Cox regression analysis indicated that the independently significant predictors of survival were diagnosis (P < .006), ECOG performance status (P < .04), number of bone metastases (P < .008), presence of visceral metastases (P < .03), hemoglobin count (P < .009), and survival estimate (P < .00005). Diagnosis, ECOG performance status, and visceral metastases covaried with surgeon survival estimate. Linear regression and receiver-operator characteristic assessment confirmed that clinician estimation was the most accurate predictor of survival, followed by hemoglobin count, number of visceral metastases, ECOG performance status, primary diagnosis, and number of bone metastases. Nevertheless, survival estimate was accurate in predicting actual survival in only 33 (18%) of 181 patients.
CONCLUSION: A better means of prognostication is needed. In this article, we present a sliding scale for this purpose.
INTRODUCTION
Decisions regarding potential surgery for metastatic disease require reliable data about patient survival and quality of life. This investigation evaluated factors that have been cited as important correlates to patient postoperative survival and compared them to surgeon preoperative survival estimates.
This information is sought for many reasons. It helps to set appropriate expectations for the patient, family, and medical staff. Data about cost, risk, and quality of life are often conflicting, but if these data could be weighed properly, it would help to define the most appropriate treatment for patients with metastatic bone disease.1-7 Utility analysis is an effective method to evaluate complex problems like this and can guide decision making by the individual patient, other stake holders, and society in general. It depends on the accurate prediction of survival, among other factors. Unfortunately, little data exist to inform these end-of-life decisions.8-18
The present study was designed to investigate the prognosis of patients with metastatic bone disease presenting to the orthopedic surgeon in need of skeletal surgical stabilization. Often the decision to operate on a patient with metastatic bone disease is strongly influenced by the expected survival of the patient.1,2,4-7,17,18 Previous studies focusing on specific primary cancers have been conducted and remain valuable tools in this regard.5,6,10,19-21 However, these studies do not provide a general approach that may be applied to this clinical scenario.
Much of the orthopedic literature on prognostication is devoted to the definition of clinical, radiologic, and biologic markers of disease in the nonmetastatic patient population. It is assumed that the same parameters are applicable across all stages of disease.17,18,21-26 The patients investigated in this series are, by definition, incurable, and many are in the terminal stage of their disease. We sought to investigate some well-recognized prognostic parameters to assess their value in determining the survival of patients who require treatment of metastatic disease to the bone. In addition, we sought to investigate how well these parameters stacked up against the clinical judgment of the surgeon and to translate this clinical evaluation into a reproducible format.11,17,18,27,28
METHODS
Between September 22, 1999 and March 13, 2003, all patients who were operated on for pathologic fracture or related indications were entered into a prospective quality-control database. Survival data were determined at two time points (April 2003 and May 2004). Approval for the use of clinical material was provided by the institutional review board.
Clinical parameters acquired included date of birth, sex, primary diagnosis, site of surgery, use of systemic therapy, and Eastern Cooperative Oncology Group (ECOG) performance status. Age was divided for survivorship analysis into patients older than 65 years and patients 65 years old and younger. For methodologic purposes, the primary diagnosis was grouped into five broad categories comprising lung cancer, breast cancer, prostate cancer, renal cancer, and other cancers. These categories represented the most prevalent of the diseases in this series. Site of surgery was classified into the upper extremity, the lower extremity, and the spine. Systemic therapy in the context of this study includes all forms of nonsurgical and nonradiotherapeutic adjuvants (eg, cytotoxic, hormonal, immunologic, metabolic, and so on) administered before or after surgery. The ECOG performance status29 was used as a clinical indicator of general functional condition. The ECOG performance status definitions are as follows: 0 for normal function, 1 for minimal functional impairment, 2 for impairment amounting to spending less than 50% of time in bed, 3 for impairment amounting to spending more than 50% of time in bed, and 4 for being completely bed bound. The performance status was measured at baseline as part of the standard evaluation of all patients on our service.
The number of bone metastases was based on radiologic assessment and bone scans. For statistical assessment of survivorship analysis, metastasis was categorized into patients with a single metastatic focus and patients who had multiple bone metastases. The presence of visceral metastases was based on routine computed tomography scans of the chest, abdomen, and pelvis. Nodal involvement was categorized for methodologic purposes into presence or absence of adenopathy and was assessed by physical examination and imaging studies of thoracic, abdominal, and pelvic cavities. Nodal involvement alone, independent of visceral involvement, was not classified as visceral metastases per se.
Blood investigations were based on the routine preoperative assessment for surgery. Laboratory data acquired included serum albumin and calcium, hemoglobin, and lymphocyte counts.10,21-27 For purposes of survivorship analysis, the following data were divided into two groups: serum albumin, 3.5 g/dL v greater than 3.5 g/dL; serum calcium, 9 mg/dL v greater than 9 mg/dL; hemoglobin, 10 g/dL v greater than 10 g/dL; and lymphocyte count, 500 cells/μL v greater than 500 cells/μL.
The authors' service represents a specialized orthopaedic oncology unit in a tertiary referral cancer center. Most patients were treated by the senior author (P.J.B.) who has close to 30 years of experience in the field. All official consults are maintained in a department-owned database. This excludes informal consults for patients who nevertheless may have had surgery and were included in this study. At the time of first consult, all patients deemed suitable for surgical stabilization were reviewed by the surgeon who made an assessment of the expected survival of the patient in months. This was recorded, and the patient was subsequently observed. By definition, the prediction was accurate if the actual survival was within 20% of the surgeon estimate of expected survival.
Data on the occurrence of death was gleaned from an institution-owned patient management database (Disease Management System version 5.2, 1996; Memorial Sloan-Kettering Cancer Center, New York, NY). Data was maintained in hard copy and electronically in a spreadsheet in Microsoft Excel version 10 for Windows NT (Microsoft, Redmond, WA). Unless otherwise stated, clinical parameters are presented as the mean ± standard deviation, and survival in months is presented as the median with 95% CIs.
Statistical analysis was performed using SPSS version 11.5 for Windows NT (SPSS Inc, Chicago, IL). Univariate analysis using the Kaplan-Meier method with log-rank assessment was performed to assess prognostic significance of individual risk factors (Figs 1 to 6). Cox regression multivariate analysis of factors found statistically significant by univariate analysis was used to assess for the independent prognostic value of risk factors. Comparisons of variables defined by means and standard deviations alone were performed with the Student's two-tailed t test. Linear regression analysis and receiver-operator characteristic (ROC) curve assessment of independently prognostic variables were used to show the relative value of these factors in predicting survival (Fig 7). For the parameter of primary diagnosis, the hazard ratio was derived from survival analysis and used as the independent variable in linear regression and ROC assessment. Statistical significance was defined as P < .05.
Data from the parameters found to be statistically significant by multivariate analysis were consolidated in a sliding scale based on median survival and 95% CI (Fig 8). To use the sliding scale, the relevant parameters were placed on the line representing the median and 95% CI in order of chronologic clinical assessment (ie, diagnosis, ECOG performance status, bone metastasis, visceral metastasis, and hemoglobin). When there was disparity between factors (eg, a patient with low hemoglobin but single metastasis), the relative position of the other factors and their statistical significance as well as the clinician estimate influenced placement.
RESULTS
Between September 22, 1999 and March 13, 2003, 191 patients were entered onto the study. All patients underwent surgical procedures pertaining to metastatic bone disease. The data were periodically updated during the 4 years of the study. The time of last accrual of data was May 19, 2004. At the time of last accrual, 50 patients remained alive, and there had been 141 deaths. Survival estimates had been provided for 181 patients. Kaplan-Meier and Cox regression analyses were performed on all 191 patients. Overall, median survival in the cohort was 8 months (95% CI, 5.4 to 10.6 months).
There were 86 female patients (45%) and 105 male patients (55%) in this series. Age was evenly distributed between the two groups (study population: mean age, 61.8 ± 13.6 years; mean for males, 61.6 ± 13.5 years; mean for females, 62.0 ± 13.7 years). The most prevalent diagnostic groups included 39 patients (20%) with lung cancer, 37 patients (19%) with breast cancer, 31 patients (16%) with renal cancer, 15 patients (8%) with prostate cancer, and 69 patients (36%) with other cancers (Table 1).
Within the stipulated time, there had been about 320 official fracture-related consults on the service. Therefore, the operative rate was 60% (191 of 320 consults). Surgery was performed on the spine in 34 patients (18%), on the upper extremity in 52 patients (27%), and on the lower extremity in 105 patients (55%). The main indications for surgery were fractures or impending fractures of bones of the extremity afflicted by metastatic disease. Impending fractures in the extremities were assessed, as per published guidelines, to be at high risk for fracture and prophylactically treated if they were osteolytic more than osteoblastic, were in the peritrochanteric area of the femur rather than in non–weight-bearing bones, involved more than half of the circumference of a bone, and were associated with functional pain.9 Joint replacement surgery was performed for fractures in the hip, knee, shoulder, and elbow. Fracture stabilization using plates, screws, and rods supplemented with cement was performed in long bones of the extremity. Spinal surgical decompression and stabilization were performed for neurologic compromise or potential neurologic compromise in the spine, especially in radiation-resistant tumors. Amputations were performed in patients with otherwise unmanageable limbs who were unsuitable for extensive surgery. Accordingly, joint replacement surgery was performed in 96 patients (50%), extremity fracture stabilization was performed in 59 patients (31%), spinal surgical decompression and stabilization were performed in 34 patients (18%), and amputations were performed in two patients (1%).
Twenty patients required reoperation. In general, this was a result of disease progression, which compromised the previous construct. Two patients required surgery for infection. Four patients had surgery for disease progression in the form of excision of soft tissue masses, which did not compromise previous constructs. Fracture stabilization failure was the most common indication for reoperation (seven patients). Other indications included five joint replacement failures and two spinal instrumentation failures.
During the course of the study, 85 patients (45%) did not receive systemic therapy, and 106 patients (55%) did receive systemic therapy. ECOG performance status was favorable (0, 1, or 2) in 93 patients (49%) and unfavorable (3 and 4) in 93 patients (49%). ECOG performance status was not available in five patients. Bone metastasis was single in 55 patients (29%) and multiple in 136 patients (71%). Nonskeletal visceral metastases were present in 113 patients (59%) and absent in 78 patients (41%). Nodal involvement was noted in 34 patients (18%) and absent in 157 patients (82%). The mean albumin level in the cohort was 3.9 ± 0.6 g/dL. The mean serum calcium measured 9.2 ± 1.0 mg/dL. The mean hemoglobin level was 11.5 ± 1.9 g/dL, and the mean absolute lymphocyte count was 1,500 ± 5,100 cells/μL.
Of the clinical parameters, only primary diagnosis, use of systemic therapy, and ECOG performance status were statistically significant predictors of survival by Kaplan-Meier analysis. As defined, age older than or 65 years was not found to be statistically significant for predicting survival (P < .8). Sex (P < .4) and site of surgery (P < .8) were similarly unremarkable for predicting survival.
The primary histologic type was significant in predicting survival (P < .008). Lung cancer patients fared the worst, with a median survival (Table 2) from the time of orthopedic consult of 4 months (95% CI, 2.2 to 5.7 months). Renal cell carcinoma patients fared the best, with a median survival time of 20 months (95% CI, 15.2 to 24.8 months). The median survivals of all other patients were clustered between the values for these two histologic types (Fig 1).
The use of systemic therapy (Fig 2) was a significant predictor of survival (P < .0004). Systemic therapy use was associated with a poorer prognosis, with a median survival time of 5 months (95% CI, 3.3 to 6.7 months). Paradoxically, patients who did not receive systemic therapy had better prognoses (median survival time, 16 months; 95% CI, 5.7 to 26.3 months).
ECOG performance status (Fig 3) was significantly predictive of survival (P < .0001). Median survival time for patients with an ECOG performance status of 0, 1, or 2 was 14 months (95% CI, 8.1 to 19.9 months), which contrasted strongly with patients with an ECOG performance status of 3 or 4 (median survival time, 5 months; 95% CI, 2.8 to 7.2 months).
Both the presence of visceral metastases and the number of bones involved by metastatic disease were significant predictors of survival. Patients with single bony metastasis (Fig 4) did significantly better than patients with multiple bony metastases (P < .00001). Median survival time was 24.7 months (95% CI, 17.5 to 32 months) for patients with single bony metastasis compared with 6 months (95% CI, 3.6 to 8.4 months) for patients with multiple bony metastases. Nodal involvement was not a significant predictor of survival (P < .08). Patients with visceral metastases (Fig 5) did significantly worse than patients without visceral metastases (P < .00001), with median survival times of 6 months (95% CI, 4 to 8 months) versus 24.7 months (95% CI, 15.2 to 34 months), respectively.
Several of the blood test results correlated with survival. Hemoglobin (Fig 6A) was a significant predictor of survival (P < .0001). Patients with a low hemoglobin level had a median survival time of 3 months (95% CI, 1.3 to 4.7 months), whereas patients with higher levels survived longer (median survival time, 10.3 months; 95% CI, 5.5 to 15.2 months). Albumin level (Fig 6B) had a significant prognostic value (P < .004). Patients with low albumin had a median survival time of 5.6 months (95% CI, 2.6 to 8.6 months), whereas patients with normal albumin fared better (median survival time, 10.3 months; 95% CI, 5 to 15.7 months). Absolute lymphocyte counts (Fig 6C) were significantly predictive of survival (P < .0003). The median survival time in patients with low lymphocyte counts was 3.2 months (95% CI, 2.3 to 4.1 months) compared with 10.3 months (95% CI, 6.4 to 14.3 months) in patients with higher counts. Serum calcium was not a significant predictor of survival (P < .8).
The parameters found to be significant by Kaplan-Meier univariate analysis were assessed for independent prognostic value using Cox regression multivariate analysis. The results of the analysis for the parameters of primary diagnosis, systemic therapy, ECOG performance status, bone metastases, visceral metastases, hemoglobin level, serum albumin level, and absolute lymphocyte count are listed in Table 2. In addition, estimated survival was included in a second multivariate analysis.
The six parameters found to be independently significant in predicting survival were primary diagnosis (P < .006), ECOG performance status (P < .04), number of bone metastases (P < .008), presence of visceral metastases (P < .03), hemoglobin level (P < .009), and survival estimate (P < .00005). Survival estimate was found to be a covariate with diagnosis, ECOG performance status, and visceral metastases. The trend 1 year later was similar, except that multiplicity of bone metastases became less significant (P < .05) and presence of visceral metastases became much more significant (P < .00003).
These six parameters were assessed using linear regression and ROC curve assessment (Fig 7). Linear regression analysis (graphs not shown) revealed a squared correlation value (R2) that was highest for estimated survival (R2 = 0.33), followed by hemoglobin (R2 = 0.15), visceral metastases (R2 = 0.11), ECOG performance status (R2 = 0.06), primary diagnosis (R2 = 0.05), and bone metastases (R2 = 0.05). Linear regression analysis (Fig 7A) of estimated survival within 20% of the actual estimate reduced the R2 value to 0.30, suggesting that the clinician's estimate could not be further optimized by simply increasing its tolerance. ROC curve assessment (Fig 7B) of survival as the state variable showed that the area under the curve and, thus, prognostic value were greatest for estimated survival, followed by visceral metastases, hemoglobin level, bone metastases, diagnosis, and ECOG performance status.
At final analysis, the survival estimate with a 20% margin of error was accurate in predicting actual survival in 33 (18%) of 181 patients. Of the remaining 148 patients, 78 (43%) were underestimated and 70 (39%) were overestimated in terms of duration of expected survival. There did not seem to be a systematic error in over- or underestimating survival. From a practical standpoint, it is often necessary to give a minimum survival estimate (ie, the minimum duration that the patient is at least likely to survive). Analysis with a 20% margin of error showed that the survival estimate was able to predict the minimum survival of 111 (61%) of 181 patients. There was no tendency towards over- or underestimation in men versus women or in patients older versus younger than 65 years. When corrected for diagnosis, the numbers in each comparison group were considerably reduced. In the major diagnostic groups, however, there was no significant trend towards over- or underestimation with respect to sex or age.
When the patients were categorized into groups of survivors, namely short (survival for 3 months), intermediate (survival > 3 and 9 months), and long (> 9 months), the data indicates that, although the clinician estimate is the most accurate prognostic parameter, it is apparently miscalibrated. Defined in this manner, 34 (64%) of 53 short survivors, 22 (53%) of 41 intermediate survivors, and 56 (64%) of 87 long survivors were categorized accurately (Fig 7A). In the short survival group, actual survival (1.4 ± 1.1 months) was significantly lower (P < .000001) than the surgeon estimate (4.5 ± 3.9 months). In the intermediate survival group, actual survival (5.9 ± 1.7 months) was also significantly lower than the surgeon estimate, which was 8.1 ± 5.7 months (P < .02). In the long survival group, actual survival (24.9 ± 10.4 months) was significantly longer than the surgeon estimate, which was 14.8 ± 9.3 months (P < .000001). Hence, short and intermediate survivors were overestimated and long survivors were underestimated in terms of expected duration of survival.
By considering all significant factors listed in Table 2, a sliding scale may be constructed (Fig 8). To test this scale, five subsequent patients not in this study were gleaned from a department-owned database, and their parameters were compared with the chart (Table 3). Figure 8B illustrates two patients from Table 3 mapped onto the chart as bands. Patient 1, who had a diagnosis in the other category and an ECOG performance status of 3, had visceral metastases, multiple bony metastases, and a hemoglobin level of more than 10 g/dL. The intersection of all these values was ultimately dictated by the upper limit of the range representing ECOG performance status and the lower limit of the range representing the hemoglobin level, yielding an estimated survival of 6 to 7 months. The actual survival was 7 months. Patient 3, who had a diagnosis in the other category and an ECOG performance status of 4, had visceral metastases, multiple bony metastases, and a hemoglobin level of less than 10 g/dL. In contrast, the intersection of all these values corresponded to the upper limit of the range representing the hemoglobin level and the lower limit of the range representing the presence of visceral metastases. This yielded a survival estimate of 4 months, which corresponded to the actual survival. As shown from these results, patients with poor prognostic factors are reasonably predicted to have poor survival. Patients who die prematurely from nononcologic underlying medical problems and patients with discordant prognostic factors are less accurately predicted, although this difference is small (2 months in patient 4 in our example). In patients with discordant factors, the clinician estimate is used as an additional parameter to guide placement on the scale.
DISCUSSION
This article attempts to answer a difficult question. What is the expected survival in patients presenting with metastatic bone disease? In their early description of treatment for this condition, Beals et al8 stated that the indication for surgery was that these "fractures were predictable." The idea that there should be a reasonable life expectancy before considering surgery is a relatively recent suggestion.1,2,4-7 In addition, being able to predict the expected survival of a patient has fundamental implications on the kind of reconstruction used. Patients with a prolonged life expectancy should receive an appropriately durable reconstruction, as opposed to the individual with a short life expectancy for whom an expedient method may be preferable.3,10
The present study has one major limitation; all patients considered for surgery had been physician selected to be in an ill-defined group with an expected survival reasonable enough to benefit from surgery. Also, issues of comorbidity are undoubtedly important but were not considered in our study. Recent work suggests that these issues should receive closer scrutiny.30 Hence, this data may not be applicable to patients who are not deemed reasonable surgical candidates or who refused surgery. Terminally ill patients with less than a 1-month prognosis were not operated on because they had a prohibitive operative risk. The main value of this analysis is in predicting survival in all other patients who were at least surgical candidates at the outset, which is the more common clinical scenario for the orthopedic surgeon.11,27,28
The main parameters assessed were gleaned from previous studies that have looked at prognostic factors in a number of conditions.10,19-22,25,31-33 The clinical parameters of age and sex were not expected to be of prognostic significance because these two parameters were evenly distributed among the good and bad performers. Perhaps surprisingly, the site of surgery was not predictive of survival.3 This has interesting implications. Spine surgery, for example, has been regarded as high-risk surgery that should be justified by an adequate estimated survival.7 Our findings suggest that this concept is misleading. The other interpretation of this data is that the successful surgical treatment of the patient's metastatic disease may improve survival to the point that the actual region involved becomes immaterial.34
Primary site was found to be a significant predictor of survival. The four most prevalent diagnostic groups (lung, breast, renal cell, and prostate cancer) are representative of most series on the subject.9,21 Patients with lung cancer fared the worst, and patients with renal cancer fared the best, as shown in Figure 1 and Table 2. Diagnosis was found to be independently predictive of survival on multivariate analysis.17,18
Systemic therapy use was found to be a poor prognostic indicator for survival. This should not be taken to mean that systemic therapy is contraindicated in this condition. Because of the complex combinations of different forms of systemic therapy in patients with prostate, breast, and lung cancer, it is difficult to interpret the meaning of this result. In this series, systemic therapy use was unevenly distributed between the primary diagnostic categories. Of the four common diagnoses, systemic therapy was predominantly used in breast and prostate patients. In lung cancer patients and all other patients (the patients with the worst prognoses), systemic therapy was used in approximately half of the patients. The overwhelming majority of renal cancer patients (the patients with the best survival) did not receive systemic therapy. This selection bias accounted for the poor prognosis associated with systemic therapy.
The ECOG performance status was found to be a significant prognostic factor. This has similarly been found to be a useful prognostic indicator in other studies. Furthermore, it's simplicity of use and reproducibility makes it a valuable index for prognostication.29,35-37
Bone metastases were found to be independently predictive of survival.4 The concept of polyostotic metastases being poorly prognostic has important implications. At least part of the morbidity and mortality that result from pathologic fractures is the predisposition to infirmity and immobility,34 which are readily amenable to improvement by surgical intervention. This would be a strong justification for surgical intervention in these patients as a means of prolonging life and not just palliation alone.
Hematologic markers have been used in various studies as prognostic indicators.21-27 Surprisingly, serum calcium was not found to be a significant predictor of survival both by univariate Kaplan-Meier and t test analyses. Hypercalcemia is a problem that is actively pursued and treated in this institution; hence, it may not have become a clinically apparent problem. This is supported by the absence of hypercalcemia seen in this series; the levels in survivors (9.27 ± 0.86 mg/dL) were not significantly different from levels in nonsurvivors (9.23 ± 1.03 mg/dL). Historically, this factor may have been important, but modern oncologic practice including bisphosphonate treatment may have neutralized this prognostic factor. Serum albumin, lymphocyte counts, and hemoglobin level were all found to be significant predictors of survival. However, only hemoglobin level was found to be an independent predictor of survival.17,18,22-25,27
Diagnosis, ECOG performance status, number of bone metastases, presence of visceral metastases, hemoglobin level, and survival estimate were independent predictors of survival.4-6,10,17-19,27,36,37 The trend 1 year later was similar, except that multiplicity of bone metastases became less significant (P < .05) and presence of visceral metastases became much more significant (P < .00003); this is likely a reflection of disease progression and increased tumor load. In general, it was found that patients with an unfavorable primary diagnosis, poor ECOG performance status, multiple bone and visceral metastases, and low hemoglobin level survived half as long as patients with favorable parameters (Fig 8). Linear regression (Fig 7A) and ROC curve assessment (Fig 7B) suggest that the predictive values of these parameters for survival were low, accounting for between 5% and 15% of the variance in the data (R2 = 0.05 to 0.15). This weak correlation implies that these parameters are not as clinically useful as commonly assumed. The attending surgeon's prediction (R2 = 0.33) was clearly superior to any of these parameters; this may be attributed to their extensive experience in the field.11,27,28 Conversely, by considering hemoglobin level (R2 = 0.15), visceral metastases (R2 = 0.11), ECOG performance status (R2 = 0.06), primary diagnosis (R2 = 0.05), and bone metastases (R2 = 0.05), 42% (the sum of R2) predictability of survival is possible from an objective standpoint. This may be superior to that offered by experience alone.
Interestingly, within specific survival groups, the numbers of short and intermediate survivors were overestimated and the number of long survivors was underestimated in terms of duration of survival. This provides some insight into conflicting literature that, for the most part, shows that survival estimates tend to be overoptimistic but has occasionally reported the reverse. Furthermore, it has been shown that, when clinical prediction of survival increased, variability in actual survival also increased.11 The accuracy of survival estimation for the surgical patients in this series illustrates this point well.
One possible explanation in this study for why diagnosis, ECOG performance status, and visceral metastasis were all significant predictors rendered insignificant by multivariate analysis is that these parameters covaried with the surgeon's estimate or that the surgeon already considered all these factors at some level in deriving a conclusion. Yet the surgeon's estimate alone was inaccurate as a predictor of survival. Perhaps the determinant is an ambiguous combination of factors that has, to date, not been determined. It may be that the patient's cognitive state, demeanor, drive, general state, or other factors together with the more traditional parameters should all be considered in the prognostication of these patients.11,18,27,28,38,39
The suggestion that the clinician estimate is accurate but miscalibrated has been described.11,18 This may be capitalized on by getting the treating clinician to suggest a prognostic category (eg, short, intermediate, or long survival) and support this estimate using the data presented in Table 2 and Figure 8. It would prove to be a more objective undertaking than suggesting an actual number, which is rarely helpful. Such a model of assessment would be a useful adjunct to the graphical system proposed. Graphical methods of prognostication are admittedly cumbersome.4,9,11,27,28,35,40,41 The proposed sliding scale system has the benefit of allowing for the influence of the clinician estimate, which, as we and others have described, is the most accurate tool in prognostication.11,17,18,28 We believe that the clinician estimate should be considered as one of a few important criteria rather than as a unique criterion in choosing therapeutic interventions.38,39 To that end, a future study will be conducted to validate this prognostic model.
In summary, this study shows that the only independent predictors of survival in the patient with bone metastases are diagnosis, ECOG performance status, number of bone metastases, presence of visceral metastases, and hemoglobin level. These parameters distinguish, with an accuracy of 5% to 15%, the ability of patients to survive less than 6 months when these parameters are unfavorable and more than 12 months when they are favorable. An assessment by a senior surgeon independent of these factors is far more accurate at 33%. This is a wake-up call to the community at large. Justification of surgery on the basis of survival prognostication is dangerously inaccurate. There is a need to create a more accurate prognostic index for patients undergoing orthopedic surgery for bone metastases.
Authors' Disclosures of Potential Conflicts of Interest
Although all authors completed the disclosure declaration, the following author or immediate family members indicated a financial interest. No conflict exists for drugs or devices used in a study if they are not being evaluated as part of the investigation. For a detailed description of the disclosure categories, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration and the Disclosures of Potential Conflicts of Interest section in Information for Contributors.
NOTES
Supported by grants from the Biomet Oncology Fellowship and the Pearlman Limb Preservation Fund.
Presented in part at the 12th International Symposium of Limb Salvage, Rio de Janeiro, Brazil, September 15-16, 2003.
This work is original and solely owned by the authors and their institution.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Harrington KD: Introduction, in: Orthopaedic Management of Metastatic Bone Disease (ed 1). St Louis, MO, Mosby, 1988, pp 1-14
Coleman RE, Rubens RD: Bone metastases, in: Abeloff MD, Armitage JO, Lichter AS, et al (eds): Clinical Oncology (ed 2). New York, NY, Churchill-Livingstone, 2004, pp 836-871
Bohm P, Huber J: The surgical treatment of bony metastases of the spine and limbs. J Bone Joint Surg Br 84:521-529, 2002
Tokuhashi Y, Matsuzaki H, Toriyama S, et al: Scoring system for the preoperative evaluation of metastatic spine tumor prognosis. Spine 15:1110-1113, 1990
James JJ, Evans AJ, Pinder SE, et al: Bone metastases from breast carcinoma: Histopathological-radiological correlations and prognostic features. Br J Cancer 89:660-665, 2003
Pietropaoli MP, Damron TA, Vermont AI: Bone metastases from squamous cell carcinoma of the head and neck. J Surg Oncol 75:136-141, 2000
Cohen DB, Riley LH: Management of metastatic carcinoma to the spine: Surgical treatment, in: Sim FH, Menendez LR (eds): Orthopedic Knowledge Update: Musculoskeletal Tumors (ed 1). Rosemont, IL, American Academy of Orthopedic Surgeons, 2002, pp 331-348
Beals RK, Lawton GD, Snell WE: Prophylactic internal fixation of the femur in metastatic breast cancer. Cancer 28:1350-1354, 1971
Mirels H: Metastatic disease in long bones: A proposed scoring system for diagnosing impending pathologic fractures. Clin Orthop Relat Res 249:256-264, 1989
Wedin R: Surgical treatment for pathologic fracture. Acta Orthop Scand Suppl 72:1-29, 2001
Glare P, Virik K, Jones M, et al: A systematic review of physicians' survival predictions in terminally ill cancer patients. BMJ 327:195, 2003
Fidler M: Prophylactic internal fixation of secondary neoplastic deposits in long bones. Br Med J 1:341-343, 1973
The Breast Specialty Group of the British Association of Surgical Oncology: British Association of Surgical Oncology Guidelines: The management of metastatic bone disease in the United Kingdom. Eur J Surg Oncol 25:3-23, 1999
Zhou Z, Redaelli A, Johnell O, et al: A retrospective analysis of health care costs for bone fractures in women with early-stage breast carcinoma. Cancer 100:507-517, 2004
Hansen BH, Keller J, Laitinen M, et al: The Scandinavian Sarcoma Group Skeletal Metastasis Register: Survival after surgery for bone metastases in the pelvis and extremities. Acta Orthop Scand Suppl 75:11-15, 2004
Miner TJ, Jaques DP, Shriver CD: A prospective evaluation of patients undergoing surgery for the palliation of an advanced malignancy. Ann Surg Oncol 9:696-703, 2002
Vigano A, Bruera E, Jhangri GS, et al: Clinical survival predictors in patients with advanced cancer. Arch Intern Med 160:861-868, 2000
Vigano A, Dorgan M, Buckingham J, et al: Survival prediction in terminal cancer patients: A systematic review of the medical literature. Palliat Med 14:363-374, 2000
Yamashita K, Koyama H, Inaji H: Prognostic significance of bone metastasis from breast cancer. Clin Orthop Relat Res 312:89-94, 1995
Rana A, Chisholm GD, Khan M, et al: Patterns of bone metastasis and their prognostic significance in patients with carcinoma of the prostate. Br J Urol 72:933-936, 1993
Coleman RE: Skeletal complications of malignancy. Cancer 80:1588-1594, 1997
Conlan DP: Value of lymphocyte counts as a prognostic index of survival following femoral neck fractures. Injury 20:352-354, 1989
Carson JL, Terrin ML, Barton FB, et al: A pilot randomized trial comparing symptomatic vs. hemoglobin-level-driven red blood cell transfusions following hip fracture. Transfusion 38:522-529, 1998
Gruson KI, Aharonoff GB, Egol KA, et al: The relationship between admission hemoglobin level and outcome after hip fracture. J Orthop Trauma 16:39-44, 2002
Koval KJ, Maurer SG, Su ET, et al: The effects of nutritional status on outcome after hip fracture. J Orthop Trauma 13:164-169, 1999
Walls J, Bundred N, Howell A: Hypercalcemia and bone resorption in malignancy. Clin Orthop Relat Res 312:51-63, 1995
Muers MF, Shevlin P, Brown J: Prognosis in lung cancer: Physicians' opinions compared with outcome and a predictive model. Thorax 51:894-902, 1996
Vigano A, Dorgan M, Bruera E, et al: The relative accuracy of the clinical estimation of the duration of life for patients with end of life cancer. Cancer 86:170-176, 1999
Oken MM, Creech RH, Tormey DC, et al: Toxicity and response criteria of the Eastern Cooperative Oncology Group. Am J Clin Oncol 5:649-655, 1982
Piccirillo JF, Tierney RM, Costas I, et al: Prognostic importance of comorbidity in a hospital-based cancer registry. JAMA 291:2441-2447, 2004
Jensen JS, Tondevold E: Mortality after hip fractures. Acta Orthop Scand 50:161-167, 1979
Kyo T, Takaoka K, Ono K: Femoral neck fracture: Factors related to ambulation and prognosis. Clin Orthop Relat Res 292:215-222, 1993
Townsend PW, Rosenthal HG, Smalley SR, et al: Impact of postoperative radiation therapy and other perioperative factors on outcome after orthopedic stabilization of impending or pathologic fractures due to metastatic disease. J Clin Oncol 12:2345-2350, 1994
Nozue M, Oshiro Y, Kurata M, et al: Treatment and prognosis in colorectal cancer patients with bone metastasis. Oncol Rep 9:109-112, 2002
Blay JY, Lasset C, Carrie C, et al: Multivariate analysis of prognostic factors in patients with non HIV-related primary cerebral lymphoma: A proposal for a prognostic scoring. Br J Cancer 67:1136-1141, 1993
Masutani M, Tsujino I, Fujie T, et al: Moderate dose-intensive chemotherapy for patients with non-small cell lung cancer: Randomized trial, can it improve survival of patients with good performance status? Oncol Rep 6:1045-1050, 1999
Mani S, Todd MB, Katz K, et al: Prognostic factors for survival in patients with metastatic renal cancer treated with biological response modifiers. J Urol 154:35-40, 1995
Pirovano M, Maltoni M, Nanni O, et al: A new palliative prognostic score: A first step for the staging of terminally ill cancer patients—Italian Multicenter and Study Group on Palliative Care. J Pain Symptom Manage 17:231-239, 1999
Maltoni M, Nanni O, Pirovano M, et al: Successful validation of the palliative prognostic score in terminally ill cancer patients: Italian Multicenter Study Group on Palliative Care. J Pain Symptom Manage 17:240-247, 1999
Culine S, Kramar A, Saghatchian M, et al: Development and validation of a prognostic model to predict the length of survival in patients with carcinomas of an unknown primary site. J Clin Oncol 20:4679-4683, 2002
Hess KR, Abbruzzese MC, Lenzi R, et al: Classification and regression tree analysis of 1000 consecutive patients with unknown primary carcinoma. Clin Cancer Res 5:3403-3410, 1999(Saminathan S. Nathan, Joh)