A New Prognostic Index and Comparison to Three Other Indices for Patients With Brain Metastases: An Analysis of 1,960 Patients in the RTOG Database

Purpose

The purpose of this study is to introduce a new prognostic index for patients with brain metastases and compare it with three published indices. Treatment for brain metastases varies widely. A sound prognostic index is thus important to guide both clinical decision making and outcomes research.

Methods and Materials

A new index was developed because of limitations in the three existing indices and new data (Radiation Therapy Oncology Group 9508) are available since the others were developed. All four indices were compared using the Radiation Therapy Oncology Group database of 1,960 patients with brain metastases from five randomized trials. The ability of the four indices to distinguish its separate classes was determined statistically. Advantages and disadvantages of each index are discussed.

Results

Recursive partitioning analysis (RPA) and the new Graded Prognostic Assessment (GPA) had the most statistically significant differences between classes (p < 0.001 for all classes).

Conclusions

The new index, the GPA, is as prognostic as the RPA and more prognostic than the other indices. The GPA is the least subjective, most quantitative and easiest to use of the four indices. Future clinical trials should compare the GPA with the RPA to prospectively validate these findings.

Introduction

Brain metastases are a common problem in clinical oncology today. Brain metastases are diagnosed in more than 170,000 patients in the United States each year (1). Historically, the prognosis was uniformly poor, with survival in the 2–4-month range (2). In recent years, with improving imaging and such treatments as stereotactic radiosurgery (SRS), it has become clear that not all patients with brain metastases have the same prognosis, and use of the same treatment for all patients is no longer appropriate.

Treatment options include surgery, whole-brain radiation therapy (WBRT), SRS, or some combination thereof. Randomized trials now guide our decision making to some degree. Patchell et al.(3) showed that surgery and WBRT were better treatment than WBRT alone in patients with a resectable solitary brain metastasis. Radiation Therapy Oncology Group (RTOG) Protocol 9508 showed a survival advantage for patients undergoing WBRT plus SRS compared with WBRT alone for those with solitary brain metastases, as well as improved performance status and decreased steroid dependence. RTOG 9508 also showed a trend toward improved survival in patients with two to three metastases (4).

Despite this progress, clinicians are still faced with a common dilemma: which treatment is most appropriate for a given patient? A useful prognostic index could guide these treatment decisions, as well as future research. All three previously published indices 5, 6, 7, 8 have limitations. The three indices include: (1) the RTOG recursive partitioning analysis (RPA) 5, 6, with three classes: I (patients < 65 years, Karnofsky Performance Status [KPS] ≥ 70, controlled primary tumor, and no extracranial metastases), III (KPS < 70), and II (all patients not in Class I or III; Table 1); (2) the Score Index for Radiosurgery (SIR) (7), which is the sum of scores (0–2) for each of five prognostic factors (age, KPS, status of systemic disease, number of lesions, and largest lesion volume; Table 2); and (3) the Basic Score for Brain Metastases (BSBM) (8), which is the sum of scores (0–1) for three prognostic factors (KPS, control of primary tumor, and extracranial metastases (Table 3).

The limitations of the previous indices include: (1) the RPA and BSBM do not incorporate number of metastases; (2) all three previous indices require the estimation of control of systemic disease, which is fraught with inconsistency because of variation in type and timing of imaging tests, and (3) the SIR requires treatment factors (volume of the largest lesion at the time of radiosurgery) when the point of prognostic indices is to predict outcome before, not after, treatment decisions are made, thus guiding the decisions.

The new index, the Graded Prognostic Assessment (GPA), was developed because of these limitations and because new data (RTOG 9508) are now available. The GPA is the sum of scores (0, 0.5, and 1.0) for four factors. Components of the GPA are age, KPS, extracranial metastases (none and present), and number of metastases (one, two to three, and more than three). The rationale for the development of this index was three-fold: (1) the need in incorporate new data from RTOG 9508 that showed the number of metastases to be prognostic; (2) the desire to eliminate components in the other indices that are difficult to quantify and/or subjective, such as control of extracranial disease; and (3) to eliminate treatment factors because the point of a prognostic index is to guide treatment choice, rather than reflect treatment results (Table 4).

In addition to introducing the new GPA index, another purpose of this study is to determine which of the four indices is most prognostic, using data from the RTOG database. It is important to distinguish prognostic from predictive factors. A prognostic factor identifies good vs. bad outcome irrespective of treatment used, whereas a predictive factor identifies good vs. bad outcome for a specific treatment. Thus, an accurate prognostic index can distinguish classes of patients by prognosis before treatment, thus informing treatment choice, rather than reflecting treatment result. An improved prognostic index also would guide the design of future clinical trials.