OBJECTIVES: To examine the evidence used by the Centers for Medicare and Medicaid Services (CMS) to update coverage decisions for items/services within the Coverage with Evidence Development (CED) program.

METHODS: Using CMS.gov, we identified items/services with a National Coverage Determination (NCD) requiring CED with a coverage update. We characterized evidence referenced in updated decision memos.

RESULTS: Of 26 items/services covered under CED from 2005-2023, 10 (38%) had an updated decision memo, with a median (IQR) of 7.9 (6.6-12.1) years between the first and subsequent decision memos. All NCD revisions were based on internal technology assessments and 3 (30%) also included external technology assessments.

Overall, 232 publications were cited in the 10 updated decision memos, with a median (IQR) of 17 (7-28) publications per decision memo. There were 56 (24%) publications representing CED studies, including 13 (23%) randomized clinical trials (RCTs), 1 (2%) non-randomized trial, 5 (9%) prospective cohorts, and 37 (66%) retrospective cohorts. There were 114 (49%) publications corresponding to non-CED studies, including 21 (18%) RCTs, 1 (1%) non-randomized trial, 50 (44%) prospective cohorts, 39 (34%) retrospective cohorts, and 3 (3%) cross-sectional studies. Among these 114 publications from non-CED studies, 3 (3%) studies enrolled only U.S. Medicare beneficiaries, 42 (37%) enrolled U.S. participants (coverage unspecified), and 69 (61%) enrolled international participants. Additionally, there were 62 other cited evidence in the updated decision memos, including 33 (53%) meta-analyses/systematic reviews, 18 (29%) clinical practice guidelines, 8 (13%) professional society recommendations/consensus statements, and 3 (5%) appropriate use criteria. CMS’s pre-specified questions in all updated decision memos were addressed by citing at least one non-CED study.

CONCLUSIONS: CMS leverages CED requirements to generate clinical evidence about new items/services; however, this study found that updated decision memos more often cited publications from non-CED studies, many of which had less robust study designs and enrolled non-US participants.