文献速递流行病学与生物统计临床试验与 GCP

「个性化」「个体化」「精准」干预随机试验多样性高、透明度低且偏倚风险高

Randomized trials of "personalized," "individualized," and "precision" interventions are very diverse and have low transparency and high bias.
PubMed 文献 · J Clin Epidemiol · 2026-05-16

该研究调查了2020至2022年MEDLINE上标题中使用「个性化」「个体化」或「精准」描述干预措施的随机对照试验,共纳入262项。结果显示这些术语被 interchangeable 地用于以非基因组干预为主的异质性干预,大多数试验结论有利但透明度低、偏倚风险高。

要点

  • 「个性化」「个体化」「精准」在RCT中互换使用,多指非基因组干预。
  • 多数试验结论对个性化干预有利(70.6%),但68.6%总体偏倚风险高。
  • 数据和代码共享率极低(分别为5.0%和0.4%)。

核心结论

  • 当代RCT中「个性化」「个体化」「精准」标签基本互换使用,涵盖异质性大且以非基因组为主的干预。
  • 多数试验结论有利,但透明度低且偏倚风险高。

研究设计与样本

  • 调查MEDLINE中2020至2022年标题使用「个性化」「个体化」或「精准」描述干预的RCT,共纳入262项。
  • 收集试验特征、干预特征、个体化定制、透明度指标和偏倚风险数据。

主要发现

  • 「个性化」使用最多(n=129,49.2%),其次「个体化」(n=120,45.8%)和「精准」(n=13,5.0%)。
  • 多数试验比较个性化干预与非个性化对照组(n=225,85.9%),聚焦治疗应用(n=186,71.0%)。
  • 个性化最常基于生活方式、心理特征或疾病分类,遗传和组学特征较少见。
  • 多数试验实施单一干预(n=177,67.6%),为每位参与者定制(n=217,82.8%),常通过个体化剂量或内容(n=109,41.6%)。
  • 在221项比较个性化与非个性化对照的RCT中,摘要结论有利者156项(70.6%),混合36项(16.3%),不利29项(13.1%)。

关键数值与推荐

  • 数据共享仅13项(5.0%),代码共享仅1项(0.4%)。
  • 大多数试验(n=162,68.6%)总体偏倚风险高。

临床怎么用

  • 临床解读「个性化」「个体化」「精准」干预RCT时需警惕术语异质性,不应望文生义。
  • 应关注试验的透明度指标和偏倚风险,谨慎采信有利结论。

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 AND OBJECTIVES: The terms "personalized," "individualized," and "precision" are increasingly used to describe interventions, yet their operational meaning is often unclear. Despite extensive conceptual debate, no empirical work has examined how these terms are applied in randomized controlled trials (RCTs) and how transparent and reliable the results of these trials are. METHODS: We surveyed MEDLINE for RCTs published between 2020 and 2022 that used the terms "personalized," "individualized," or "precision" in their title to describe an intervention. We collected data on trial characteristics and conclusions, intervention features, individual-level tailoring, transparency indicators, and risk of bias. Interventions were classified according to the number of personalized components, modes of tailoring, and dosage variability. RESULTS: A total of 262 RCTs were included. The term "personalized" was used most frequently (n = 129, 49.2%), followed by "individualized" (n = 120, 45.8%) and "precision" (n = 13, 5.0%). Most trials compared personalized interventions versus nonpersonalized control groups (n = 225, 85.9%) and focused on therapeutic applications (n = 186, 71.0%). Behavioral, digital, and medication-based interventions predominated. Personalization was most often based on lifestyle, psychological characteristics, or disease classification, whereas genetic and omics-based features were relatively uncommon. The majority of trials implemented a single intervention (n = 177, 67.6%) tailored differently for each participant (n = 217, 82.8%), often through individualized dosage or content (n = 109, 41.6%). Of the 221 RCTs comparing personalized interventions versus nonpersonalized controls, abstract conclusions were favorable to the personalized intervention in 156 (70.6%), mixed in 36 (16.3%), and unfavorable in 29 (13.1%). Few meaningful differences were observed across studies according to the term used. Transparency indicators such as data (n = 13, 5.0%) and code sharing (n = 1, 0.4%) were rare, and most trials (n = 162, 68.6%) were judged to be at high overall risk of bias. CONCLUSION: In contemporary RCTs, the labels "personalized," "individualized," and "precision" are applied largely interchangeably to a wide range of heterogeneous interventions that are predominantly nongenomic. Most of these trials have favorable conclusions, low transparency, and high risk of bias.

来源:J Clin Epidemiol,查看原文。版权归原作者所有。

引用

  1. Russo L, Lentini N, Soru L, et al. Randomized trials of "personalized," "individualized," and "precision" interventions are very diverse and have low transparency and high bias. J Clin Epidemiol. 2026. PMID: 42144086. doi:10.1016/j.jclinepi.2026.112338

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。