Quality of life: An independent prognostic variable in a general population of cancer patients receiving chemotherapy

Abstract

19596 Background: It is uncertain whether the current Quality of Life (QL) instruments can facilitate clinical decision-making in oncology. Since the most important outcome in oncology is survival, we investigated QL domains from 2 contrasting QL instruments, EORTC QLQ-C30 and the Quality of Life Index (QLI) to determine if these instruments can partition a heterogeneous population of cancer patients into distinct prognostic groupings. Methods: 1,200 patients consented to participate in Cancer Treatment Center of America's QL program between Mar ‘01 and May ’05. We identified 494 deaths in this group. Most common tumors were breast (n=319), colon (n=159), and lung (n=209). Among 434 patients undergoing definitive treatment 159 had stage 4 and 49 had stage 3 disease, the other patients had failed definitive therapy. The median survival was 39 weeks for newly diagnosed patients and 29.9 weeks for patients with recurrent disease. We used cutpoint analysis using statistical strength of association with QL scores to determine the association between QL score and survival. This information was then used in a recursive partitioning analysis to group our cancer patient population into mutually exclusive prognostic groups. Results: We found several QL domains had a non-linear relationship with survival, which indicates that a relatively small change in QL score could result in a large change in survival. QLI Health/Functioning domain, had the strongest association with survival. Subsequent iterations of the recursive partitioning found that a combination of clinical and QL domain scores partitioned the patient population into distinct groups, whose median survival ranged from 13–100 weeks. Patients with the best prognosis had high QLI health satisfaction scores health and had hormonally dependent tumors. The patients with the poorest survival had intermediate QLI health satisfaction scores, poor appetite, and poor emotional function. Conclusion: This retrospective analysis found specific combination of QL domains and scores that partitioned patients with advanced cancer into distinct prognostic groupings. This analysis indicates that our current instruments have the potential to be developed into tools that could facilitate clinical decision-making. No significant financial relationships to disclose.