Comparing survival of older ovarian cancer patients treated with neoadjuvant chemotherapy versus primary...

Area of Science:

• Oncology
• Health Services Research
• Biostatistics

Background:

• Accurate comorbidity assessment is crucial for ovarian cancer patient care and research.
• Existing indices like the Charlson Comorbidity Index (CCI) may not fully capture comorbidity burden in ovarian cancer.
• Health services research requires reliable tools to adjust for baseline health status in comparative effectiveness studies.

Purpose of the Study:

• To develop and validate a Multidimensional Comorbidity Index (MCI) for ovarian cancer.
• To compare the accuracy of the MCI against the CCI in predicting 1-year mortality.
• To evaluate the MCI's utility in reducing confounding in comparative effectiveness research, specifically regarding neoadjuvant chemotherapy (NACT).

Main Methods:

• Utilized SEER-Medicare data for patients with stage IIIC/IV ovarian cancer (2010-2015).
• Employed partial least squares regression, a machine learning algorithm, to develop the MCI using pre-diagnosis claims and 1-year mortality.
• Assessed MCI and CCI performance (discrimination and calibration) for 1-year mortality and evaluated confounding reduction in NACT vs. primary surgery analysis.

Main Results:

• The MCI showed superior discrimination for 1-year mortality (c-index: 0.75) compared to the CCI (c-index: 0.59).
• MCI demonstrated better calibration between predicted and observed mortality risk.
• Controlling for MCI, not CCI, eliminated the significant association between NACT and increased 1-year mortality.

Conclusions:

• The MCI is a more accurate predictor of 1-year mortality in ovarian cancer patients than the CCI.
• The MCI effectively reduces confounding in comparative effectiveness research, particularly for treatments like NACT.
• The MCI offers an improved tool for health services research involving ovarian cancer patients.