From March 1997 through September 1998, a streamlined endpoint review and closure process was followed. Three changes were implemented for this period: 1) a single reviewer adjudicated all cases independently, unless questions arose; 2) for all cancers other than lung cancer, only pathology reports were required to confirm the endpoint; and 3) only the death certificate was required to confirm cause of death. The process was further streamlined in October 1998, allowing for review by endpoint specialists from the CARET Coordinating Center staff and a single reviewer from the ERC.
The final change during the active follow-up phase came in October 2003 when, as a cost-saving measure, we eliminated the collection and review of pathology reports and began relying solely on self-reports of cancer. Our decision to make this change was based on an evaluation of the accuracy of participant self-report of cancer, especially lung cancer, our primary endpoint. We examined the adjudication outcome of participant reported cancers and found that 91% of cases reported as lung cancer were closed as such. Most of the discrepancies in reporting were due to reporting recurrences or reporting metastases as new primaries. The confirmation rates were high for other commonly reported cancers as well, including prostate (96%), breast (94%), bladder (97%), and colorectal (88%). These findings assured us that self-report of cancer was an acceptable procedure. Based on this evaluation, we made several modifications to the data collection questionnaire to address the reasons for self-reported diagnoses with the review of medical records. In the majority of cases, this was concerned with differentiating a new primary cancer from recurrent, metastatic, and benign diseases. Thus, actual rates of agreement during the self-reporting period (October 2003 - June 2005) are likely higher than we observed in our evaluation.
State cancer registry and National Death Index (NDI) linkages were conducted in 2015 to extend follow up for endpoints. NDI linkage was complete through 2013; state cancer registry linkage was through 2012 for WA and CA, and through 2013 for CT. Cancer registries of the five states in which CARET enrollments occurred were targeted for the linkage, with approval obtained for three registries: California Cancer Registry, Connecticut Tumor Registry, and Washington State Cancer Registry. Regulatory guidelines in Maryland and Oregon at present prohibit access to cancer registry data for the aims of the current CARET grant. Of the 18,314 total CARET participants, 12,927 (71%) were enrolled by a study center located in CA, CT, or WA; and another 1,109 (6%) enrolled by other study centers had a history of residence in one of the three states. In total, records from 14,036 (77%) participants were included in the cancer registry linkages. Probabilistic software was used to match CARET records to cancer registry records based on data elements common to both repositories, including full social security number, name, sex, race, and date of birth (which are complete for 99% of CARET participants). Linkage results were classified as matches to previously reported CARET endpoints or new endpoints based on degree of matching between common data elements (e.g., date of diagnosis, cancer site, and histologic type) and manual review of pathology reports and other medical records collected during active follow up. The registry data were also used to adjudicate cases closed previously based on self-report only.
NDI linkage was performed to ascertain vital status and underlying cause of death on a total of 13,830 participants who were alive at the end of active follow-up or for whom report of death had not been confirmed by death certificate. Matches were determined on the basis of the NDI probabilistic score and classification group assignment, electronic evaluation of CARET data elements not incorporated in the NDI algorithms (e.g., date of last contact), and manual review of records. Discrepancies in cause of death and cancer diagnoses between NDI and cancer registry records were reviewed as part of the adjudication process.
• Number of deceased participants by cause of death and intervention arm assignment (PDF)
• Number of CARET Prostate Cancer Primaries by Stage of Disease (PDF)
• Number of CARET Breast Cancer Primaries by Stage of Disease (PDF)
• Number of CARET Colorectal Cancer Primaries by Stage of Disease (PDF)