Article type
Year
Abstract
Background: Previous studies have shown that randomised clinical trials (RCTs) with inadequate methodological quality exaggerate intervention efficacy significantly, and that many RCTs are designed with a significant risk of overlooking important intervention benefits. Objective: To study the reported methodological quality and sample size of all RCTs published in 'Gastroenterology' as original articles from 1964 through 1999 and to explore whether these aspects differed within different disease areas and whether there was any improvement in the methodological quality during the study period.
Methods: Electronic searches and handsearches were combined. The therapeutic area of the RCT was registered. Diseases with <10 RCTs were categorised as miscellaneous. Methodological quality focusing on the adequacy of the reported randomisation, double-blinding, and follow up, was assessed by separate components and a validated dichotomised composite 5-point scale (Jadad et al., Controlled Clinical Trials 1996; 17:1-12). Associations between methodological quality, sample size, disease, and year of publication were analysed by Kruskal-Wallis non-parametric analysis of variance.
Results: In total 385 RCTs were identified. The number of RCTs dealing with different diseases is shown in Fig. 1. Of all RCTs, 160 (42%) reported adequate generation of allocation sequence, 150 (39%) reported adequate allocation concealment (26% by central randomisation and 13% by sealed envelopes), and 240 trials (62%) were adequately double blinded. A total of 289 RCTs (75%) were classed as high quality trials (>2 points) according to the composite scale. The median number of patients per intervention arm was 23 (interquartile range 10 to 50; range 1 to 1,107). The methodological quality differed significantly within diseases regarding the generation of allocation sequence (p=0.004), allocation concealment (p=0.000) (Fig. 2), double blinding (p=0.001), and quality score (p =0.001). The sample size was not significantly different in the therapeutic areas (p=0.152). During the study period there was a significant improvement in the proportion of RCTs with adequate generation of allocation sequence (p= 0.004), adequate allocation concealment (p=0.008), and with high quality score (p=0.005), but not in the proportion of double blinded trials (p=0.142). The median number of patients per intervention arm increased significantly during the study period (p=0.006).
Conclusions: Methodological quality of RCTs published in 'Gastroenterology' varies significantly according to disease area. In addition, a significant improvement of quality and number of patients was observed during the study period 1964-1999.
Methods: Electronic searches and handsearches were combined. The therapeutic area of the RCT was registered. Diseases with <10 RCTs were categorised as miscellaneous. Methodological quality focusing on the adequacy of the reported randomisation, double-blinding, and follow up, was assessed by separate components and a validated dichotomised composite 5-point scale (Jadad et al., Controlled Clinical Trials 1996; 17:1-12). Associations between methodological quality, sample size, disease, and year of publication were analysed by Kruskal-Wallis non-parametric analysis of variance.
Results: In total 385 RCTs were identified. The number of RCTs dealing with different diseases is shown in Fig. 1. Of all RCTs, 160 (42%) reported adequate generation of allocation sequence, 150 (39%) reported adequate allocation concealment (26% by central randomisation and 13% by sealed envelopes), and 240 trials (62%) were adequately double blinded. A total of 289 RCTs (75%) were classed as high quality trials (>2 points) according to the composite scale. The median number of patients per intervention arm was 23 (interquartile range 10 to 50; range 1 to 1,107). The methodological quality differed significantly within diseases regarding the generation of allocation sequence (p=0.004), allocation concealment (p=0.000) (Fig. 2), double blinding (p=0.001), and quality score (p =0.001). The sample size was not significantly different in the therapeutic areas (p=0.152). During the study period there was a significant improvement in the proportion of RCTs with adequate generation of allocation sequence (p= 0.004), adequate allocation concealment (p=0.008), and with high quality score (p=0.005), but not in the proportion of double blinded trials (p=0.142). The median number of patients per intervention arm increased significantly during the study period (p=0.006).
Conclusions: Methodological quality of RCTs published in 'Gastroenterology' varies significantly according to disease area. In addition, a significant improvement of quality and number of patients was observed during the study period 1964-1999.