The natural history of most rare diseases is incompletely understood and usually relies on studies with low level of evidence. Consistent with the goals for future research of rare disease research set by the International Rare Diseases Research Consortium in 2017, the purpose of this paper is to review the recently developed method of quantitative retrospective natural history modeling (QUARNAM) and to illustrate its usefulness through didactically selected analyses examples in an overall population of 849 patients worldwide with seven (ultra‐) rare neurogenetic disorders. A quantitative understanding of the natural history of the disease is fundamental for the development of specific interventions and counseling afflicted families. QUARNAM has a similar relationship to a published case study as a meta‐analysis has to an individual published study. QUARNAM relies on sophisticated statistical analyses of published case reports focusing on four research questions: How long does it take to make the diagnosis? How long do patients live? Which factors predict disease severity (eg, genotypes, signs/symptoms, biomarkers)? Where can patients be recruited for studies? Useful statistical techniques include Kaplan‐Meier estimates, cluster analysis, regression techniques, binary decisions trees, word clouds, and geographic mapping. In comparison to other natural history study methods (prospective studies or retrospective studies such as chart reviews), QUARNAM can provide fast information on hard clinical endpoints (ie, survival, diagnostic delay) with a lower effort. The choice of method for a particular drug development program may be driven by the research question and may encompass combinatory approaches.

中文翻译：

---

孤儿药开发的定量回顾性自然史模型

大多数罕见疾病的自然史尚不完全清楚，通常依赖于证据水平较低的研究。与国际罕见病研究联盟 2017 年制定的罕见病研究未来研究目标一致，本文的目的是回顾最近开发的定量回顾性自然历史模型（QUARNAM）方法，并通过教学法说明其实用性。选定的分析示例来自全球 849 名患有七种（超）罕见神经遗传疾病的患者。对疾病自然史的定量理解对于特定干预措施的发展和对患病家庭的咨询至关重要。QUARNAM 与已发表的案例研究的关系类似于元分析与已发表的个别研究的关系。QUARNAM 依赖于针对四个研究问题的已发表病例报告的复杂统计分析：做出诊断需要多长时间？患者能活多久？哪些因素可以预测疾病的严重程度（例如，基因型、体征/症状、生物标志物）？在哪里可以招募患者进行研究？有用的统计技术包括 Kaplan-Meier 估计、聚类分析、回归技术、二元决策树、词云和地理映射。与其他自然史研究方法（前瞻性研究或回顾性研究，如图表审查）相比，QUARNAM 可以以较低的工作量提供关于硬临床终点（即生存、诊断延迟）的快速信息。