Article type
Year
Abstract
Objective: The randomized controlled trial (RCT) represents the highest level of evidence used to assess an intervention. An RCT's ability to influence clinical practice depends on: 1) its size, 2) the outcomes measured, and 3) the methodologic quality of the trial. To assess these factors in RCTs involving NIDDM patients we performed two independent searches of the MEDLINE database looking for RCTs published between January 1994 and November 1995.
Methods: Information relating to trial design was independently extracted by two investigators. We retrieved 195 unique citations of which 103 were excluded as not being RCTs. Of the remaining 92 RCTs 79 were available for analysis.
Results: Forty-two percent of RCTs were published in either Diabetic Medicine or Diabetes Care. The median number of patients randomized was 40 (range 5 to 2769). Median duration of study was 22 weeks (.2-260). Glycemic control was repotted as an outcome measure in all but 4 RCTs (95%). The next most frequently reported outcomes were lipid levels (63% of RCTs), body weight (56% of RCTs), and a measure of insulin secretion and/or insulin action (44% of RCTs). Only two studies (2.5%) reported major clinical endpoints and five studies (6.3%) reported a formal measure of quality of life. Ninety-one percent of patients randomized had complete follow-up. The method of randomization was described in only 15% of RCTs. In the 46 RCTs which were described as double-blind, an appropriate method of blinding was reported in only 21 RCTs (46%).
Discussion: Our data suggest a need for more large RCTs with major clinical endpoints and a greater awareness of patient related outcomes (quality of life). The development of a valid instrument to assess the quality of NIDDM RCTs is needed.
Methods: Information relating to trial design was independently extracted by two investigators. We retrieved 195 unique citations of which 103 were excluded as not being RCTs. Of the remaining 92 RCTs 79 were available for analysis.
Results: Forty-two percent of RCTs were published in either Diabetic Medicine or Diabetes Care. The median number of patients randomized was 40 (range 5 to 2769). Median duration of study was 22 weeks (.2-260). Glycemic control was repotted as an outcome measure in all but 4 RCTs (95%). The next most frequently reported outcomes were lipid levels (63% of RCTs), body weight (56% of RCTs), and a measure of insulin secretion and/or insulin action (44% of RCTs). Only two studies (2.5%) reported major clinical endpoints and five studies (6.3%) reported a formal measure of quality of life. Ninety-one percent of patients randomized had complete follow-up. The method of randomization was described in only 15% of RCTs. In the 46 RCTs which were described as double-blind, an appropriate method of blinding was reported in only 21 RCTs (46%).
Discussion: Our data suggest a need for more large RCTs with major clinical endpoints and a greater awareness of patient related outcomes (quality of life). The development of a valid instrument to assess the quality of NIDDM RCTs is needed.