Lessons Learned Regarding Missing Clinical Stage in the National Cancer Database

Area of Science:

• Oncology
• Biostatistics
• Health Informatics

Background:

• The National Cancer Database (NCDB) is a key resource for analyzing national cancer treatment patterns.
• Data abstraction issues between 2004-2007 led to substantial missing clinical stage information in the NCDB.
• This missingness poses a significant challenge for accurate breast cancer research, particularly for patients receiving neoadjuvant chemotherapy (NAC).

Purpose of the Study:

• To investigate the impact of missing clinical stage data on breast cancer studies.
• To assess the bias introduced in analyzing trends of neoadjuvant chemotherapy (NAC) versus adjuvant chemotherapy (AC) due to missing stage information.
• To evaluate the effectiveness of multiple imputation in correcting these biases.

Main Methods:

• Examined the percentage of missing clinical stage data in the NCDB from 2004-2013.
• Compared trends in NAC use with and without accounting for missing clinical stage data.
• Employed multiple imputation using the chained equations approach with predictive mean matching to address missing data.

Main Results:

• Clinical stage was missing in 56% of cases from 2004-2007, compared to 12% from 2008-2013.
• Missing clinical stage was more prevalent in adjuvant chemotherapy (AC) patients than in neoadjuvant chemotherapy (NAC) patients (31% vs. 12% overall).
• Multiple imputation revealed significant increases in NAC use across all stages from 2004-2013, contrasting with analyses excluding missing data.

Conclusions:

• Data abstraction rules from 2004-2007 created substantial missing clinical stage data in the NCDB, potentially biasing breast cancer research.
• Excluding cases with missing clinical stage or the years 2004-2007 can lead to inaccurate conclusions regarding NAC trends.
• Multiple imputation is a viable strategy to mitigate bias caused by missing clinical stage data in the NCDB.