Cancer Staging Over Time Fields Database

Over the years, the criteria/definitions for the collection of staging data has changed. A majority of those adjustments are due to changes in the American Joint Committee on Cancer (AJCC) staging manuals, which change every 8-10 years. The Surveillance Research Program (SRP) that oversees the Surveillance, Epidemiology and End Results (SEER) registries, has updated how stage is collected based on the AJCC changes and these updates are reflected in the Research and Research Plus data. These changes can cause differences in the stage distribution, survival and trends over time.

Researchers have been requesting to look at stage over time (covering multiple AJCC editions), but until now, this capability has either been limited, or very complex to implement.

In 2019, the Staging Over Time team at SEER started developing detailed algorithms that would modify the definitions of T, N, M, and Stage Group so that they could be consistent over time. These algorithms take into effect the changes of what is eligible for AJCC staging, based on AJCC 8th edition.

For the current databases, the following schemas have been completed and are ready for review:

• Bladder
• Colon/Rectum
• Liver
• Intrahepatic Bile Duct
• Stomach (incl C160)

Although SEER staging data is based on AJCC definitions, SEER provides a combined T, N, M and Stage group based on all information available for the patient but does not provide data for specific timeframes (clinical, pathological, neoadjuvant).

For completing the SEER data items related to staging (Extent of Disease (EOD) 1988-2003, 2018+, and Collaborative Stage (CS) 2004-2015), pathological information generally takes priority over clinical information. However, when neoadjuvant therapy is done (followed by surgical resection), the most extensive stage is recorded (which is usually the clinical stage).

For 2016-2017, Union for International Cancer Control (UICC) 7th edition was collected by the SEER registries as SEER transitioned from CS back to EOD. Both clinical and pathological timeframes were collected and then a "combined" T, N, M and Stage were derived and made available in SEER*Stat.

Although all schemas follow similar algorithms for staging over time, each schema will have detailed information available on specific changes.

Data available through the Cancer Staging Over Time Fields database

For each of the schemas where AJCC is applicable, the following information will be available for each of the timeframes (long term and short term):

1. EOD Primary Tumor
2. EOD Regional Nodes
3. EOD Mets
4. Summary Stage 2018
5. Derived T
6. Derived N
7. Derived M
8. Derived Stage Group

Timeframes for schemas

There are two major timeframes available: 1992+ and 2004+. Years available are dependent on when the applicable AJCC chapter was implemented, or due to changes between editions that made going back in time more difficult. Each schema will have detailed information on the years available and which AJCC edition is being used over time. Note that data available for 1992+ may be based on a different edition of AJCC compared to data available for 2004+.

1992+ - "Long term fields":

• Due to how stage data was collected prior to 2004, there are some limitations with this data. A separate mets field was not available prior to 2004. If a patient had organ metastasis, that information was collected in EOD extension and the derivation was a TX, M1. If a patient had distant lymph nodes, that information was collected in EOD Lymph Nodes and the derivation was a NX, M1.
• Starting in 2004, a separate Mets field was implemented so that T and N could be collected individually, along with specific metastatic information. In order to make data for 2004 consistent with data 1992+, all cases diagnosed 2004+ that had mets present were changed as follows:
  • Metastasis to organs or carcinomatosis
    • If distant metastasis (bone, brain, lung, liver, other sites, or carcinomatosis) to organs were coded, the derived T value would be converted to TX.
  • Metastasis to distant lymph nodes
    • If distant metastasis to distant lymph nodes were coded, the derived N value would be converted to NX.

2004+ - "Short term fields":

• All these cases have the separate mets at dx field. Cases with metastatic disease (organ or distant lymph nodes) will maintain their original T and N data.

Database Details

There are two Cancer Staging Over Time Fields databases available to request:

All databases are available in the Case Listing, Frequency, and Survival sessions in SEER*Stat for the November 2022 data submission. They are identical to the SEER Research Plus database other than the specialized fields.

1. Cancer Staging Over Time Fields Database
   • This one is linked to county-level attributes, which include county-level SES, rurality, and demographics.
   • There are two datasets available: one that includes all tumor records from 1992-2020 from the SEER 12 registries, and one that includes all tumor records from 2000-2020 from the SEER 17 registries.
   • Both the SEER 12 and SEER 17 datasets include the long-term and short-term stage fields described above.
2. Cancer Staging Over Time Fields with Census Tract Attributes Database
   • This includes cases based on the SEER 17 registries for cases diagnosed 2006+.
   • There are no geographic identifiers included in this database due to confidentiality concerns.
   • It does not include Alaska Native Tumor Registry data.
   • For detailed information about census tract-level SES and rurality variables, refer to Census Tract-level SES and Rurality Database.

Data Limitations and Analytical Considerations

Modifying definitions over time

For the most part, modifying definitions over time was straightforward based on how EOD and CS were defined. With the older data (1992-2003), there were times that a particular code definition included structures or lymph nodes that were separate codes in later editions. Additional algorithms were put in place to modify the data as much as feasible.

To help with consistency over time, other available data items, not typically used for staging, were used to categorize cases. For example, for EOD Regional Nodes (N), if a NX was derived and there were positive nodes, the N value was changed to match the positive nodes. In this situation, the assumption was made that the data item Regional Nodes Positive was correct. Depending on the schema, other changes like this were made to make the data more compatible over time.