Article type
Year
Abstract
Introduction: Tools available to researchers attempting to identify reports of randomized controlled trials (RCTs) in MEDLINE include: 1) the publication type "randomized controlled trial" which was introduced in 1991 and 2) the Cochrane Highly Sensitive Search Strategy (Cochrane Strategy).
Objective: Calculate the sensitivity of the publication type "RCT' and of the Cochrane Strategy for identifying reports of RCTs in four major hypertension-related journals. Calculate the incremental yield (predictive value positive) of the three phases of the Cochrane Strategy. Determine the appropriateness of excluding reports as "not randomized" based on the absence of the word "random" (or one of its variations) in the abstract of an RCT.
Methods: The Journal of Cardiovascular Pharmacology, Hypertension, the Journal of Hypertension, and the American Journal of Hypertension were retrospectively handsearched from each journal's volume 1 through the year 1996 to identify RCTs. A total of 18,041 reports were assessed. Sensitivity and yield calculations for the electronic searches used the handsearch results as a "gold standard."
Results: For the years 1990 and prior, MEDLINE properly coded 59% of identified RCTs with the publication type "RCT' while the Cochrane Strategy retrieved 93% of RCTs. For the years 1991-1996, the sensitivity values for the publication type and Cochrane Strategy were 88% and 97%, respectively. For the years 1990 and prior, 152 of 369 records (yield = 41%) retrieved by Phase 1 of the Cochrane Strategy were RCTs not coded with the publication type "RCT." Phases 2 and 3 of the Cochrane Strategy produced yields of 10% and 5% respectively. For the years 1991-1996, Phases 1, 2 and 3 produced yields of 9%, 5% and 2%, respectively, beyond the records coded with the publication type "RCT." The root term "random$" was included in the abstract of 70%, 47% and 0% of RCTs retrieved by Phases 1, 2 and 3, respectively.
Discussion: The publication type "RCT' is not sufficiently sensitive for identifying RCTs in MEDLINE for a systematic review, while the Cochrane Search is sufficiently sensitive. All three phases of the Cochrane Strategy produced a reasonable yield with the exception of Phase 3 for the years 1991-1996. The full text of controlled trials should be assessed before excluding those trials as "not randomized."
Objective: Calculate the sensitivity of the publication type "RCT' and of the Cochrane Strategy for identifying reports of RCTs in four major hypertension-related journals. Calculate the incremental yield (predictive value positive) of the three phases of the Cochrane Strategy. Determine the appropriateness of excluding reports as "not randomized" based on the absence of the word "random" (or one of its variations) in the abstract of an RCT.
Methods: The Journal of Cardiovascular Pharmacology, Hypertension, the Journal of Hypertension, and the American Journal of Hypertension were retrospectively handsearched from each journal's volume 1 through the year 1996 to identify RCTs. A total of 18,041 reports were assessed. Sensitivity and yield calculations for the electronic searches used the handsearch results as a "gold standard."
Results: For the years 1990 and prior, MEDLINE properly coded 59% of identified RCTs with the publication type "RCT' while the Cochrane Strategy retrieved 93% of RCTs. For the years 1991-1996, the sensitivity values for the publication type and Cochrane Strategy were 88% and 97%, respectively. For the years 1990 and prior, 152 of 369 records (yield = 41%) retrieved by Phase 1 of the Cochrane Strategy were RCTs not coded with the publication type "RCT." Phases 2 and 3 of the Cochrane Strategy produced yields of 10% and 5% respectively. For the years 1991-1996, Phases 1, 2 and 3 produced yields of 9%, 5% and 2%, respectively, beyond the records coded with the publication type "RCT." The root term "random$" was included in the abstract of 70%, 47% and 0% of RCTs retrieved by Phases 1, 2 and 3, respectively.
Discussion: The publication type "RCT' is not sufficiently sensitive for identifying RCTs in MEDLINE for a systematic review, while the Cochrane Search is sufficiently sensitive. All three phases of the Cochrane Strategy produced a reasonable yield with the exception of Phase 3 for the years 1991-1996. The full text of controlled trials should be assessed before excluding those trials as "not randomized."