"Multiple Primary Rules" by NCI/SEER to count incident cancer cases

[cashaowan]

I'm working on NPCR data on cancer incidence. It turned out I accidentally count more cancer events than what CDC/SEER rules would. After some extensive search, what happened is that sequence 00 is correct for the person’s only primary, however, sequence 01 could, for example, include both a first lung primary and also the first colon primary for the same person, which CDC sometimes excludes or combines under site-specific rules.

In other words, by including all “01” in cancer count, I double-(or triple-) charging a personal who got two separate primaries, whereas CDC’s rule set may only one of those two as an “incident” for the site/time window selected. For example, if a person had a lung primary in 2015 and a colon primary in 2017, my code will add 2 to cancer count, but CDC’s “2015-2020 lung + colon rate” might only count 1 (only the first lung, for example) depending on the site-specific multiple primary rules.

I am trying to identify a table referred to as “Multiple Primary (MP)” decision table, which will help decide whether to count as a primary cancer occurrence. So far hasn't been able to. Anyone knows how to identify this table and incorporate it as a Marco in SAS programming?

[Kathryn_SAS]

Here are some links to papers that you may be able to use:

https://pmc.ncbi.nlm.nih.gov/articles/PMC3638182/ 

https://www.researchgate.net/publication/229438287_Identifying_primary_and_recurrent_cancers_using_a...

If you are looking for basic counts, then you may be able to use first.byvar logic in a DATA step to add a count variable that is 1 for "sequence 00" and 0 for "sequence 01".  However, you would need to include some sample data and show what you are expecting to get to determine if that is an option.

[cashaowan]

This is super helpful! I think, for "sequence 01", it is not just simply bypass because this needs to be scrutinized by the multiple primary rules to decide. Sometimes it will be counted; other times, it will not be counted, depending on specific sites and time window identified. I will look into the paper and let you know if more questions.