Boschetti Laura, Nilius Henning, Ten Cate Hugo, Wuillemin Walter A, Faes Livia, Bossuyt Patrick M, Bachmann Lucas M, Nagler Michael
Department of Clinical Chemistry, Inselspital University Hospital, University of Bern, Bern, Switzerland.
Division of Haematology, Cantonal Hospital of Lucerne, Lucerne, Switzerland.
Front Cardiovasc Med. 2024 Nov 29;11:1420000. doi: 10.3389/fcvm.2024.1420000. eCollection 2024.
Early detection and diagnosis of venous thromboembolism are vital for effective treatment. To what extent methodological shortcomings exist in studies of diagnostic tests and whether this affects published test performance is unknown.
We aimed to assess the methodological quality of studies evaluating diagnostic tests for venous thromboembolic diseases and quantify the direction and impact of design characteristics on diagnostic performance.
We conducted a literature search using Medline and Embase databases for systematic reviews summarizing diagnostic accuracy studies for five target disorders associated with venous thromboembolism. The following data were extracted for each primary study: methodological characteristics, the risk of bias scored by the QUADAS QUADAS-2 instrument, and numbers of true-positives, true-negatives, false-positives, and false-negatives. In a meta-analysis, we compared diagnostic accuracy measures from studies unlikely to be biased with those likely to be biased.
Eighty-five systematic reviews comprising 1'818 primary studies were included. Adequate quality assessment tools were used in 43 systematic reviews only (51%). The risk of bias was estimated to be low for all items in 23% of the primary studies. A high or unclear risk of bias in particular domains of the QUADAS/QUADAS-2 tool was associated with marked differences in the reported sensitivity and specificity.
Significant limitations in the methodological quality of studies assessing diagnostic tests for venous thromboembolic disorders exist, and studies at risk of bias are unlikely to report valid estimates of test performance. Established guidelines for evaluation of diagnostic tests should be more systematically adopted.
PROSPERO (CRD 42021264912).
静脉血栓栓塞症的早期检测和诊断对于有效治疗至关重要。诊断试验研究中存在多大程度的方法学缺陷，以及这是否会影响已发表的试验性能尚不清楚。
我们旨在评估评估静脉血栓栓塞性疾病诊断试验的研究的方法学质量，并量化设计特征对诊断性能的方向和影响。
我们使用Medline和Embase数据库进行文献检索，以获取总结与静脉血栓栓塞相关的五种目标疾病的诊断准确性研究的系统评价。从每项原始研究中提取以下数据：方法学特征、由QUADAS-2工具评分的偏倚风险，以及真阳性、真阴性、假阳性和假阴性的数量。在一项荟萃分析中，我们比较了不太可能有偏倚的研究与可能有偏倚的研究的诊断准确性指标。
纳入了85项系统评价，包括1818项原始研究。仅43项系统评价（51%）使用了充分的质量评估工具。23%的原始研究中所有项目的偏倚风险估计为低。QUADAS/QUADAS-2工具特定领域的高偏倚风险或不明确偏倚风险与报告的敏感性和特异性的显著差异相关。
评估静脉血栓栓塞性疾病诊断试验的研究在方法学质量上存在重大局限性，有偏倚风险的研究不太可能报告有效的试验性能估计值。应更系统地采用既定的诊断试验评估指南。
PROSPERO（CRD 42021264912）。
