app

SEER Data Errors: NCI Looks To Avoid Repeats

— Results reassuring as SEER audit nears completion

MedpageToday

In July, it was learned that officials at the National Cancer Institute (NCI) had decided to withhold prostate-specific antigen data to its Surveillance Epidemiology and End Results (SEER) database from the 2014 update because of problems in that data. Here's what has happened since.

NCI officials continue to review PSA data withheld from 2014 update to the , but they've developed a "fix" designed to prevent a repeat.

Discovered during a routine quality-control check, the errors in data reported by several SEER registries subsequently led to a comprehensive audit of reporting methods. , associate director of surveillance research at SEER, originally estimated that the audit would be completed by October.

"The review of all prostate cancer cases for 2012 was completed in October, and we are doing the analysis of the data now," Penberthy said in an email. "We hope to have the publication submitted in the next couple of months."

Preliminary findings from the audit pointed toward a much lower error rate than initially estimated (15% to 20% of all cases), and updated findings suggest an error rate of <4%, she added.

"It is definitely lower, but we are still finalizing some of the analysis and carefully reviewing the data to be sure that we have the most accurate estimate of the error rate," Penberthy said.

The questionable data arose from varied interpretation of a vague "implied decimal point" in the data-entry template used by SEER registries. In the three-digit data-entry field for PSA values, the implied decimal point falls between the second and third digit. If the actual value were 4 ng/mL, for example, the correct format on the template would be 040. If the value were entered as 004, the computer software would interpret the value as 0.4 ng/mL.

"We are implementing a routine quality check for this variable in the future, and we are using the results to develop focused training for registrars who collect the data," Penberthy said.

The data in question came from prostate cancer cases reported to SEER for 2012. SEER added PSA data in 2004, and the data were incorporated for prostate cancer staging purposes in 2010. Using the data reviewed thus far, SEER officials have developed a means to "semi-automate" review of PSA values dating back to 2004.

"We anticipate that we will have, at a minimum, the 2012 to 2013 corrected data available for the April 2016 data release for SEER," said Penberthy. "We are hoping to have all older data reviewed and corrected, if needed, soon thereafter."

SEER is the only cancer surveillance program that has taken steps to review and correct PSA data, she added. The original quality-control study suggested a potentially high error rate for PSA values across all cancer surveillance programs.

When news of the PSA reporting errors surfaced, prostate cancer researchers expressed divergent views about the impact of the questionable data on published and ongoing studies. Some foresaw minimal impact, whereas others worried about possible reverberations affecting hundreds, if not thousands, of publications and ongoing studies. The final verdict awaits the updated report from SEER.

"Things really haven't changed," said , of Vanderbilt University in Nashville, Tenn., author of a widely cited editorial expressing concern about the faulty data. "Until the NCI completes its review, there's not much more to add."

  • author['full_name']

    Charles Bankhead is senior editor for oncology and also covers urology, dermatology, and ophthalmology. He joined app in 2007.