3097 - Development and validation of a multivariable prediction model for 6-month mortality in older cancer patients: the GeriAtrIc-Tumor Score of PrEdiction for Early Death (GAIT SPEED)

Background

Cancer is a disease of elderly, however data from evidence-based-medicine are missing for therapeutic decision in this population. One of the main issues is to avoid over- and undertreatment situations. Cancer treatment decision in the elderly mainly relies on the Geriatric Assessment (GA) recommended by the International Society of Geriatric Oncology (SIOG). Based on the GA, predictive scores of early death have been developed but they remain difficult to implement in daily oncological practice. In this study, we proposed a simple score with five clinical items to predict 6-month mortality risk in older cancer patients, to guide therapeutic decision.

Methods

A total of 603 patients aged 65 and over were prospectively included in registry in a two-center cohort study that started in November 2013. The whole cohort was divided in a development subset (n = 439), and a validation subset (n = 164) We created a multivariate Cox proportional hazard model with a selection process based on the lowest Akaïke Information Criteria. A beta-coefficient point-based scoring system was used to weight each predictor. Discrimination used the survival Harrel’s C index with 95% CI. Clinical impact was assessed using decision curves.

Results

The mean age was 81.2 ± 6.1 years. Most patients were women and had locally advanced (38%) or metastatic cancers (45%). Colorectal, breast and lung cancers were the most common types. At 6 months, the mortality rate was 17.5%. The score we developed, namely GAIT-SPEED, included five clinical variables: unintentional loss weight of at least 5% of the previous year, slow gait speed < 0.8 m/s, cancer site, cancer extension, and exclusive supportive care decision. We identified 4 increasing risk group (low, medium, high and very-high), ranged from 2% to 61%. The score was well calibrated, and discrimination was good with a Harrel’s C index of 0.75 [0.69-0.81]. Up to 14% of risk of early death, decision curves encourages for a cancer-treatment.

Conclusions

In this study, we developed and validated a simple score easy to implement in daily oncological practice, to predict early death in older cancer patients and guide oncologists in their treatment decision.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

Frédéric Pamoukdjian.

Funding

Has not received any funding.

Disclosure

All authors have declared no conflicts of interest.