Quality of statistical methods reported in randomised-controlled trials of type-2 diabetes (T2DM) published in 2016

ID: 

1017

Session: 

Poster session 1 Wednesday: Evidence production and synthesis

Date: 

Wednesday 13 September 2017 - 12:30 to 14:00

Location: 

All authors in correct order:

Ooi CP1, Ho J2
1 Universiti Putra Malaysia, Malaysia
2 Penang Medical College, Malaysia
Presenting author and contact person

Presenting author:

Cheow Peng Ooi

Contact person:

Abstract text
Background: High-quality primary clinical research data from randomised-controlled trials (RCTs) is essential for clinical decision making and incorporation into systematic reviews. An essential part of a high-quality study is the use of appropriate statistical analyses that are adequately reported. Inappropriate use and inadequate reporting of statistics have been found in medical publications.

Objective: To assess the quality of statistical methods reported in RCTs of interventions for type-2 diabetes (T2DM) in a cross-sectional study.

Methods: We searched CENTRAL, MEDLINE, PUBMED and other resources for RCTs of interventions for T2DM published in 2016. We used ‘Statistical Analyses and Methods in the Published Literature’ (SAMPL) guidelines.to assess the quality of statistical reporting. Each of the 14 statistical methods reporting domains was summarised using descriptive statistics.

Results: We included 34 RCTs from 20 journals. All the articles were inadequate for one or more reporting domains (Table 1). These were the statistical procedures to modify raw data (94%); outliers (100%) and missing data (53%); as well as verification of assumptions used in statistical tests (79%). Fewer numbers failed to evaluate the model of the statistical techniques used to compare groups (47%); mention the statistical package used (41%); define smallest clinically significant effect size (38%); methods used for each analysis (29%); description of the intent of analysis (21%); and, methods used in post hoc analyses (21%). The main outcome and associated descriptive statistics (2.9%); type-1 error (8.9%); and, analytical methods for primary objective (5.9%) were most adequately reported.

Conclusion: These critical shortcomings could have adverse effects on the interpretation of primary reports (possibly affecting patient care) and limit the ability to use these studies in systematic reviews. We suggest the use of statistical analyses and methods reporting guidelines, such as SAMPL, complementing the Consolidated Standards of Reporting Trials (CONSORT) statement to improve the quality of reporting.

Attachments: