EBMVet, Via Sigismondo Trecchi 20, Cremona, Italy.
Private Practice of Surgery, Chicago, IL 60654, USA.
J Clin Epidemiol. 2018 Sep;101:73-78. doi: 10.1016/j.jclinepi.2018.05.001. Epub 2018 May 16.
The validity of outcomes of systematic reviews is highly dependent on the extent of bias in the included primary studies. This study reports the risk of bias (ROB) of primary studies included in systematic reviews of diagnostic accuracy.
All systematic reviews of diagnostic accuracy studies published in the Cochrane database in 2015 and 2016 that used the Quality Assessment of Diagnostic Accuracy Studies-2 ROB tool and reported results with the ROB figure were eligible. The primary outcome was the prevalence of "high" or "unclear" ROB scores for the four Quality Assessment of Diagnostic Accuracy Studies-2 domains: "patient selection," "index test," "reference standard," and "flow and timing".
Of 46 eligible reviews, 35 fulfilled the inclusion criteria. A total of 1045 primary studies with 4133 bias assessments were identified. Of those, 56% (2319/4133) were assessed to be at "high" or "unclear" ROB and 44% (1814/4133) at low ROB. For all domains except "flow and timing," most outcomes were scored as "high" or "unclear" ROB. A total of 47 (47/1045; 4.5%, 3.4 to 5.9%) primary studies were scored at low ROB for all domains. Older article age was significantly associated with likelihood of "high" or "unclear" ROB (odds ratio: 1.02; 95% confidence interval: 1.01 to 1.03; P < 0.001).
Systematic reviews of diagnostic accuracy are based on studies with a majority of "high" or "unclear" bias assessments. The age of the articles explained only a small part of the variability of the score assessments, therefore not justifying an a priori exclusion of older articles in systematic reviews. There is an urgent need to improve the quality of design, conduct, and reporting of diagnostic accuracy studies.
系统评价结果的有效性高度依赖于纳入的原始研究中的偏倚程度。本研究报告了纳入系统评价的诊断准确性原始研究的偏倚风险（ROB）。
所有于 2015 年和 2016 年在 Cochrane 数据库中发表的、使用了质量评估诊断准确性研究-2（QUADAS-2）工具评估 ROB 并以 ROB 图报告结果的诊断准确性系统评价均符合纳入标准。主要结局是“高”或“不明确”QUADAS-2 四个领域（患者选择、索引测试、参考标准和流程与时间）的 ROB 评分的发生率。
46 篇符合条件的综述中，有 35 篇符合纳入标准。共纳入了 1045 项原始研究，涉及 4133 次偏倚评估。其中，56%（2319/4133）被评估为“高”或“不明确”ROB，44%（1814/4133）为低 ROB。除“流程与时间”外，所有领域的大部分结局均为“高”或“不明确”ROB。仅有 47 项（47/1045；4.5%，3.4%至 5.9%）原始研究在所有领域均被评估为低 ROB。文章的年代较久远与“高”或“不明确”ROB 的可能性显著相关（比值比：1.02；95%置信区间：1.01 至 1.03；P<0.001）。
诊断准确性的系统评价基于大部分具有“高”或“不明确”偏倚评估的研究。文章年代仅能解释评分评估变异性的一小部分，因此不能事先排除系统评价中的旧文章。急需提高诊断准确性研究的设计、实施和报告质量。
