Developing a simple outpatient prognostic model for prediction of survival in patients with advanced cancer.

Abstract

e13871 Background: Predicting the life expectancy in patients with advanced cancer is important but difficult. Patients and families need a reasonably accurate estimate of survival time for social and personal reasons. Clinical prediction overestimates survival, and predictive models must be validated in different populations. We developed a simple prognostic model using readily available parameters for estimating survival in patients with advanced solid malignancy. Methods: This was a prospective observational study of 343 patients with breast or ovarian cancer who had disease progression after receiving at least two lines of systemic therapy and gastrointestinal (GI) cancer after at least one line of systemic therapy. Multivariable Cox proportional hazards analysis was used to identify factors significantly associated with overall survival and a prognostic model was created. Using risk scores derived from the model for categories in each variable a nomogram model was constructed. C-index was used to calculate the discriminative ability of the model. The degree of calibration was evaluated using the plot method. Results: Among 343 patients included in the analysis, the median age was 50 (44-58) years, the 1-year survival rate was 34% and median overall survival (OS) was 7.8 months(m), [95% confidence interval (CI) 6.4-9.1m]. Five variables were significantly associated with OS in multivariable analysis (primary site, progression-free interval after most recent systemic therapy, lymphocyte percentage, hemoglobin, and albumin levels). Each significant variable was assigned a weighted score, which was summed and plotted on a nomogram to determine survival probability at 1 year, 3 years, 5 years and 6 years. The C -index for the model was 0.676 ± 0.015. Conclusions: We developed an accurate prognostic model to predict overall survival in Indian patients with advanced-stage, previously treated breast, ovarian or GI cancers, which is derived using readily available parameters. The nomograms derived are a potentially powerful tool to communicate information about survival. Clinical trial information: CTRI/2017/04/0084026 . Factors impacting survival and the prognostic model derived. Factors Hazard Ratio 95% CI p-value Points Site of primary  Non-GI* 0  GI 1.76 1.34 – 2.31 0.000 10.0 PFI** (m)  ≥ 6 0  3-6 1.03 0.62-1.69 0.921 0.4  < 3 1.63 1.13-2.35 0.008 8.7 Lymphocyte%  ≥ 20% 0  < 20% 1.57 1.23-2.01 0.000 8.0 Albumin (gm/dL)  ≥ 3.5 0  < 3.5 1.67 1.26-2.21 0.000 9.1 Hemoglobin levels (gm/ dL)  ≥ 12 0  <12 1.38 1.02-1.86 0.036 5.7 *Refers to breast or ovarian or both. **PFI is defined as the time interval from the starting date of the most recent treatment to the date of progression on the given treatment.