Dear Editor, I am writing to share a critique about a highly cited recently published (2018) paper on the efficacy and safety of sodium-glucose cotransporter-2 inhibitors (SGLT2i) in type 1 diabetes patients [1]. In contemporary medicine, when the knowledge in this milieu is not parallelly apparent as in type 2 diabetes mellitus patients, the paper proves vital in addressing the evidence gap [1]. I thank the authors [1] for their substantial effort to address the context. I start here with some of the discrepancies between the protocol and the published paper [1, 2]. First, it is unclear if searching of the “Web of Science” database occurred as per the protocol [1, 2]. Second, in contrast to the published paper (and its supplement), the aim of testing the safety and efficacy of weekly administered SGLT2is was not prespecified [1, 2]. Lastly, as per the protocol and the title of the study, the review intended to study the safety and efficacy of SGLT2is only (as an adjunct to insulin therapy) in type 1 diabetes patients; however, data pooled from seven trials contradict this (six tested a combination of sodium-glucose cotransporter-1 inhibitor and SGLT2i [sotagliflozin] and the remaining tested the aggregate of liraglutide, dapagliflozin, and insulin). Next, while the study aimed to test the weekly administration of SGLT2is, it primarily included trials where participants received the test drug daily. Then, the incorporation of the four conference-abstract-based data in the main (quantitative) analysis is not agreeable. While the authors assumed an unclear risk of bias in these abstracts [1], it is plausibly better to avoid a risk of bias assessment when a full manuscript is unavailable. In reality, if the risk of bias is high, the summary statistic is less likely to be of any worth. A supplementary analysis may be appropriate in such a scenario. Finally, the pooled data in the metaanalysis also do not appear conceivable. For instance, regarding side effects, data was chiefly pooled for particular doses of the respective study drugs (while the paper’s aim was not dose-specific). Another illustration: in one trial, although severe hypoglycemia occurred in one of the participants in the intervention group, a placebo recipient was also accounted for the event while pooling data [3]. In light of the pitfalls discussed above, the findings of the paper necessitate a cautious interpretation and authors may consider a revision. Disclosure Statement

中文翻译：

---

对钠-葡萄糖协同转运蛋白 2 抑制剂作为 1 型糖尿病患者胰岛素治疗的辅助手段的有效性和安全性的随机临床试验的系统评价和荟萃分析的评价

亲爱的编辑，我写这封信是想对最近发表的一篇被高度引用的（2018 年）论文发表评论，该论文关于钠-葡萄糖协同转运蛋白 2 抑制剂 (SGLT2i) 在 1 型糖尿病患者中的有效性和安全性 [1]。在当代医学中，当这种环境中的知识不像 2 型糖尿病患者那样明显时，该论文证明对于解决证据差距至关重要 [1]。我感谢作者 [1] 为解决上下文所做的大量努力。我从协议和已发表论文 [1, 2] 之间的一些差异开始。首先，不清楚是否按照协议 [1, 2] 进行了“Web of Science”数据库的搜索。其次，与已发表的论文（及其补充）相比，测试每周给药的 SGLT2is 的安全性和有效性的目的不是预先指定的 [1, 2]。最后，根据研究方案和研究标题，旨在研究 SGLT2 的安全性和有效性的审查仅（作为胰岛素治疗的辅助）用于 1 型糖尿病患者；然而，来自七项试验的汇总数据与此相矛盾（六项测试了钠-葡萄糖协同转运蛋白-1 抑制剂和 SGLT2i [sotagliflozin] 的组合，其余测试了利拉鲁肽、达格列净和胰岛素的组合物）。接下来，虽然该研究旨在测试 SGLT2is 的每周给药情况，但主要包括参与者每天接受测试药物的试验。然后，将四个基于会议摘要的数据纳入主要（定量）分析是不可接受的。虽然作者在这些摘要中假设了不明确的偏倚风险 [1]，当无法获得完整手稿时，避免偏倚评估的风险似乎更好。实际上，如果偏倚风险很高，则汇总统计数据就不太可能具有任何价值。在这种情况下，补充分析可能是合适的。最后，荟萃分析中的汇总数据似乎也无法想象。例如，关于副作用，数据主要是针对各自研究药物的特定剂量汇总的（而论文的目的不是特定剂量）。另一个例子：在一项试验中，虽然干预组的一名参与者发生了严重的低血糖，但在汇总数据时，安慰剂接受者也被考虑在内[3]。鉴于上述讨论的陷阱，论文的发现需要谨慎解释，作者可能会考虑修改。