Is This "Easy-to-Use" Tool the Best Way to Predict Survival?: Commentary on an article by J.J. Willeumier, MD, et al.: "An Easy-to-Use Prognostic Model for Survival Estimation for Patients with Symptomatic Long Bone Metastases".

Abstract

Commentary Comparatively speaking, there are more tools for predicting survival in patients with spine metastases than with long bone metastases, but good tools are needed for each1-5. The authors of this study felt strongly enough about this fact that they set out to develop a survival prediction tool for patients with long bone metastases based on their earlier work in the spine. To evaluate this tool, several criteria that define a good survival prediction tool can be examined. First, the data on which it is based must be robust and supported by valid statistical analysis. Second, a truly useful survival predication tool is best if it is easy to use. Less is more here, and the variables should be easily accessible. Third, the tool should provide meaningful survival windows that contribute to clinical decision-making. Unfortunately, there are at least 3 aspects to meaningful survival windows: recovery, durability, and economics; none have been studied adequately to date. Armed with these suggested definitions of a good survival prediction tool, the current tool can be evaluated. There are several strengths. First, the currently reported tool is backed by robust statistical support. Variables were the product of a very large derivation data set (1,520 patients), the resulting tool having been tested successfully in a validation group of 250 patients. Second, it would appear to be “easy to use,” as it is based on only 3 variables. Third, for as much as is known about what we need for meaningful time prognostication, this tool does a good job at tightly separating the patients with very poor survival (median, 2.2 months), from those with poor (median, 4.6 months), fair (median, 10.5 months), and best (median, 21.9 months) survival. A fourth and somewhat unique strength of this tool is its flexibility in allowing adjustment of the clinical profile according to changes in the prognosis for specific primary diseases over time. However, there are several weaknesses with this tool. The very nature of including only orthopaedic and radiation oncology patients, while on the one hand representing an advance from other tools, excludes those managed primarily by medical oncology and confounds by indication, possibly biasing toward better prognosis than would have been observed if all patients were examined. There were also issues with accessibility of the 3 variables, as only 74% of patients had all 3 available. Only 76% had functional scoring available, and of those patients, 47% had scoring derived from chart descriptions; of the other 53%, only approximately half had a Karnofsky Performance Score (KPS) available, the remainder having World Health Organization scores instead. Translated, only 20% of the original derivation group had a KPS available. Prior studies have also shown high variability with respect to the KPS6. Beyond the study criticisms, the reported tool may simply not be so easy to use or even the easiest system available. A report examining bone metastases for the entire skeleton predicted survival with equivalent accuracy utilizing just 2 variables (KPS and primary tumor)5. Some might argue that a tool for the entire skeleton is more practical since it includes the pelvis, where procedures may require a longer recovery time than those of the extremity. Evaluation for brain metastases in this study was done almost exclusively without imaging and was based on clinical assessment. On the other hand, imaging to detect visceral metastases in this study was done in 97% of the patients. Hence, if one is to use this system, a thorough clinical examination to detect intracerebral lesions and imaging of the viscera should be done. Further, additional bone imaging should be obtained for renal cancers to stratify them within the 2 categories based on number of metastases. In summary, the proposed tool is based on a robust derivation database, is statistically valid, and derives from earlier models by this group that have utilized the same 3 variables for spine tumors, applying it now to long bones. It has unique advantages of including orthopaedic and radiation oncology patients, excludes myeloma, and is independent of laboratory results. From an orthopaedic perspective, the usefulness of tools such as this one awaits better understanding of typical recovery courses, durability, and cost-benefit analyses for internal fixation devices versus standard endoprostheses and megaprostheses. Further validation of this tool awaits testing on a larger prospective pool of patients.