文献速递医学教育与科研方法

ClinicalTrials.gov 存在的一些数据质量问题

Some data quality issues at ClinicalTrials.gov.
PubMed 文献 · Trials · 2019-06-24

该研究分析 ClinicalTrials.gov 注册的干预性试验,经筛选后考察主要研究者与责任方数据的质量。研究识别出四类错误并量化其中三类:35,121 项试验中 12% 缺失研究者姓名或角色,71,359 对姓名与角色中 17% 的姓名为非真实人员信息,35,121 项中 1,221 项(3.5%)缺失责任方标签。作者建议采取措施预防此类错误,并强调对试验注册库开展审计的重要性。

要点

  • 35,121 项试验中 12% 缺失研究者姓名或角色。
  • 71,359 对姓名与角色中 17% 的姓名为非真实人员信息。
  • 35,121 项试验中 1,221 项(3.5%)缺失责任方标签。

核心结论

  • 研究概述了 ClinicalTrials.gov 数据的四类问题并量化其中三类,同时建议未来如何预防这些错误,强调对试验注册库开展多种审计以识别记录缺陷并加以解决很重要。

研究设计与样本

  • 分析在 ClinicalTrials.gov 注册的干预性试验,初始 112,013 条记录,经筛选要求试验于 2005 年 1 月 1 日至 2014 年 12 月 31 日间启动、干预字段含药物或生物制品、在美国机构注册、并以真实人员姓名及其研究角色列为研究者。

主要发现

  • 35,121 项试验中 12% 缺失研究者姓名或角色。
  • 71,359 对姓名与角色中 17% 的姓名为非真实人员信息。
  • 识别出 19 类姓名变体,13% 的姓名变体无法通过程序解决。
  • 35,121 项试验中 1,221 项(3.5%)缺失责任方标签。
  • 部分试验列出多名主要研究者,而试验整体责任仅应由一人承担。

临床怎么用

  • 作者建议未来可采取措施预防这些错误,并认为对试验注册库开展多种审计以识别记录缺陷并加以解决很重要。

局限性

  • 原文未报告局限性。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

原文

BACKGROUND: Clinical trial registries have been established as a form of public accountability. Sponsors ought to register their trials promptly and accurately, but this is not always done. Some of the problems include non-registration of trials, registration of trials with incomplete information, and non-reporting of trial results on time. In this study we enumerate or quantify some quality issues with respect to Principal Investigator (PI) and Responsible Party data. METHODS: We analyzed interventional trials registered with ClinicalTrials.gov. Using certain selection criteria, we started with 112,013 records, and then applied further filters. The trial had to (a) start between 1 January 2005 and 31 December 2014, (b) include a "drug" or "biological" in the "intervention" field, (c) be registered with an American authority, and (d) list a real person's name as investigator and also his or her role in the study. RESULTS: We identified four categories of errors in the ClinicalTrials.gov records. First, some data were missing. The name of the investigator, or his or her role, was missing in 12% of 35,121 trials. In examining 71,359 pairs of names and roles, 17% of the "names" were found to be not those of real persons, but instead junk information. Second, there were variations in a large number of names. We identified 19 categories of variants. We determined that 13% of the names had variants that could not be resolved using a program. Third, some trials listed many PIs each, although only one such person holds overall responsibility for the trial and therefore not more than one person should be listed as PI. Fourth, in examining whether the PI's name was available as part of the Responsible Party tag, we found that in 1221 (3.5%) of 35,121 trials, the Responsible Party tag is absent. CONCLUSIONS: We have outlined four categories of problems with data hosted by ClinicalTrials.gov and have quantified three of them. We also suggest how these errors could be prevented in future. It is important to carry out various kinds of audits of trial registries, in order to identify lacunae in the records, that they be addressed.

来源:Trials,查看原文。版权归原作者所有。

引用

  1. Chaturvedi N, Mehrotra B, Kumari S, et al. Some data quality issues at ClinicalTrials.gov. Trials. 2019. PMID: 31234923. doi:10.1186/s13063-019-3408-2

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。