Despite advances in insulin therapy and glucose monitoring, achieving glycaemic control in patients with type 1 diabetes (T1D) remains a challenge.¹ Thus, some turn to non-insulin antihyperglycaemic agents as adjunct to insulin.² Sodium glucose co-transporter inhibitors (SGLTi) are an attractive option, in part because of an insulin-independent mechanism, but also because of emerging data on cardio- and nephro-protective potential.³ However, while clinical trials showed improved glycaemic endpoints,^4-6 reviewed elsewhere,⁷ concerns about side effects, particularly diabetic ketoacidosis (DKA), remain a major limitation. Risk-mitigation strategies have been proposed⁸; yet, regulatory judgement of these strategies’ sufficiency has differed. Whereas SGLTi use in T1D is approved in Japan and for adults with body mass index (BMI) >27 kg/m² in Europe, the US Food and Drug Administration (FDA) rejected similar approval requests. Information about risks outside the clinical trial setting, where risk-benefit ratio may differ, is limited.

The T1D Exchange (T1DX) registry comprises a multi-institutional consortium of clinical and patient-reported data that enables investigation of current trends in T1D. We examined SGLTi use among adult T1DX registry participants to assess potential benefits and complications in a real-world setting.

The cross-sectional analysis cohort included 204 SGLTi users (any duration) and 7124 non-users aged ≥18 years with T1D for ≥1 year. Demographics, clinical characteristics, glycaemia, and incidence of DKA and severe hypoglycaemia (SH) were compared. Data collection methodology has been described previously.¹ As all DKA data were collected at 5-year follow-up (July 2015-April 2018), all cases were reported after the FDA warning, issued 15 May 2015.

Differences in mean haemoglobin A1c (HbA1c) were analysed using linear regression. Odds ratios (OR) for meeting the American Diabetes Association target HbA1c [<7% (<53 mmol/mol)] and experiencing adverse events were calculated with logistic regression. Age, race, sex, income, insurance, education, pump use, continuous glucose monitor use, diabetes duration, glucose monitoring checks, total daily insulin, depression and BMI were identified as potential confounders using step-wise selection and adjusted for in the models.

SGLTi use was rare and reported by <3% of adults. Compared with non-users, those who used SGLTi were more likely to have older age, longer diabetes duration, white race, private insurance, use of insulin pump or continuous glucose monitor, higher BMI and use of other adjuvant diabetes medication (Table 1). SGLTi agents included canagliflozin (50%), dapagliflozin (27%) and empagliflozin (27%) (13 participants reported use of >1 SGLTi over the course of the registry period).

Mean HbA1c was similar between users and non-users (adjusted mean difference = 0.15%; 95% CI −0.02%-0.32%; p = .06). Compared with non-users, SGLTi users had significantly lower odds of meeting target HbA1c (adjusted OR = 0.53; 95% CI: 0.31-0.91; p = .02) and higher odds of DKA but this did not reach statistical significance (OR = 2.10; 95% CI: 0.73-6.03; p = .17). In a subanalysis, the odds of DKA was 3.33 times higher for SGLTi users with ≥1 year of use compared with those with <1 year of use, but did not achieve statistical significance for this limited sample (95% CI: 0.52-31.11; p = .20). The odds of SH was also higher among SGLTi users (OR = 2.66; 95% CI 1.49-4.77; p < .001).

Hence, compared with mild HbA1c reductions seen in clinical trials,^4-7 SGLTi use among T1DX participants was not associated with lower HbA1c. It is not clear why non-users were more likely to meet the HbA1c goal. It is possible that SGLTi use was preferentially recommended to patients with higher baseline HbA1c, representing selection bias.

Increased risk of DKA associated with SGLTi use in T1D has been shown in trials [relative risk = 4.49 (95% CI: 2.88-6.99)],⁷ but this large increased risk was not observed among a limited subset of T1DX participants. However, recent real-world data from Japan assessing a cohort of 11 475 patients with T1D, including 1898 (16.5%) SGLTi users indicated that DKA risk was higher in SGLTi users compared with non-users (hazard ratio 1.66, 95% CI 1.33-2.06; p < .001).⁹

The higher incidence of SH among SGLTi users in T1DX, even after adjusting for demographic differences, was unexpected. It is possible that selection bias exists; for example, those electing for SGLTi may strive for tighter control and thereby have an increased risk of hypoglycaemia. SGLTi users were more likely to use other non-insulin glucose-lowering medications, potentially suggesting these patients represent a subset that is more willing to utilize adjunctive therapies. While reasons for prescribing SGLTi were not available, no significant difference was seen in glomerular filtration rate, the degree of albuminuria or reported frequency of cardiovascular disease between the groups (Table 1).

These data represent a real-world addition to what is known from trials regarding SGLTi use in T1D. However, we strongly caution that the observational, cross-sectional study design prohibits any causative interpretation and provides robust evidence neither in favour of nor against SGLTi use in T1D. Data are limited by the small proportion (<3%) of participants with SGLTi use, possible confounders, and the level of detail gathered in the registry.¹ Important unknown factors include exact dosage, adherence, and temporal relationships between SGLTi use and incidence of adverse effects. Further research examining glycaemic outcomes and adverse effects associated with real-world SGLTi use in T1D, particularly with implementation of risk mitigation strategies, is warranted.

尽管在胰岛素治疗和血糖监测方面取得了进展，但在 1 型糖尿病 (T1D) 患者中实现血糖控制仍然是一个挑战。¹因此，一些人转向非胰岛素抗高血糖药物作为胰岛素的辅助剂。²钠葡萄糖协同转运蛋白抑制剂 (SGLTi) 是一种有吸引力的选择，部分原因在于其不依赖于胰岛素的机​​制，还在于有关心脏和肾脏保护潜力的新数据。³然而，虽然临床试验显示血糖终点有所改善，^4-6在其他地方进行了审查，但⁷对副作用的担忧，特别是糖尿病酮症酸中毒 (DKA)，仍然是一个主要限制。已提出降低风险的策略⁸; 然而，监管机构对这些策略是否充分的判断存在差异。虽然 SGLTi 在 T1D 中的使用在日本和欧洲的体重指数 (BMI) >27 kg/m ²的成年人中得到批准，但美国食品和药物管理局 (FDA) 拒绝了类似的批准请求。临床试验环境之外的风险信息有限，风险收益比可能不同。

T1D 交换 (T1DX) 注册表包含一个多机构的临床和患者报告数据联盟，可以调查 T1D 的当前趋势。我们检查了成人 T1DX 登记参与者中 SGLTi 的使用情况，以评估现实环境中的潜在益处和并发症。

横断面分析队列包括 204 名 SGLTi 用户（任何持续时间）和 7124 名年龄≥18 岁且 T1D ≥1 年的非用户。比较了人口统计学、临床特征、血糖以及 DKA 和严重低血糖 (SH) 的发生率。数据收集方法之前已经描述过。¹由于所有 DKA 数据均在 5 年随访期间（2015 年 7 月至 2018 年 4 月）收集，所有病例均在 FDA 2015 年 5 月 15 日发出警告后报告。

使用线性回归分析平均血红蛋白 A1c (HbA1c) 的差异。使用逻辑回归计算达到美国糖尿病协会目标 HbA1c [<7% (<53 mmol/mol)] 和经历不良事件的优势比 (OR)。年龄、种族、性别、收入、保险、教育、泵的使用、连续血糖监测仪的使用、糖尿病持续时间、血糖监测检查、每日总胰岛素、抑郁症和 BMI 被确定为使用逐步选择并在模型中进行调整的潜在混杂因素.

SGLTi 的使用很少见，<3% 的成年人报告。与非使用者相比，使用 SGLTi 的人更可能年龄更大、糖尿病病程更长、白人种族、私人保险、使用胰岛素泵或连续血糖监测仪、更高的 BMI 和使用其他辅助糖尿病药物（表 1） . SGLTi 药物包括卡格列净（50%）、达格列净（27%）和恩格列净（27%）（13 名参与者报告在注册期间使用了 >1 个 SGLTi）。

使用者和非使用者之间的平均 HbA1c 相似（调整后的平均差 = 0.15%；95% CI -0.02%-0.32%；p  = .06）。与非使用者相比，SGLTi 使用者达到目标 HbA1c 的几率显着降低（调整后的 OR = 0.53；95% CI：0.31-0.91；p  = .02）和更高的 DKA 几率，但这没有达到统计学意义（OR = 2.10；95% CI：0.73-6.03；p  = .17)。在一项子分析中，使用 ≥1 年的 SGLTi 用户发生 DKA 的几率是使用 <1 年的用户的 3.33 倍，但对于这个有限的样本没有达到统计学意义（95% CI：0.52-31.11；p  = .20)。SGLTi 用户的 SH 几率也较高（OR = 2.66；95% CI 1.49-4.77；p  < .001）。

因此，与临床试验中看到的 HbA1c 轻度降低相比，T1DX 参与者中使用^{4-7次 SGLTi 与降低 HbA1c 无关。}目前尚不清楚为什么非使用者更有可能达到 HbA1c 目标。可能会优先推荐基线 HbA1c 较高的患者使用 SGLTi，这代表了选择偏倚。

试验表明，在 T1D 中使用 SGLTi 会增加 DKA 的风险 [相对风险 = 4.49 (95% CI: 2.88-6.99)]，⁷但在有限的 T1DX 参与者中没有观察到这种大幅增加的风险。然而，最近来自日本的真实世界数据评估了 11 475 名 T1D 患者队列，其中包括 1898 名（16.5%）SGLTi 用户，表明与非用户相比，SGLTi 用户的 DKA 风险更高（风险比 1.66，95% CI 1.33 -2.06；p  < .001)。⁹

即使在调整了人口统计学差异之后，T1DX 中 SGLTi 用户的 SH 发生率也更高，这是出乎意料的。可能存在选择偏差；例如，那些选择 SGLTi 的人可能会争取更严格的控制，从而增加低血糖的风险。SGLTi 使用者更有可能使用其他非胰岛素降糖药物，这可能表明这些患者代表了更愿意使用辅助疗法的子集。虽然没有提供 SGLTi 处方的原因，但两组之间的肾小球滤过率、蛋白尿程度或心血管疾病报告的频率没有显着差异（表 1）。

这些数据代表了对 T1D 中 SGLTi 使用试验已知的真实世界的补充。然而，我们强烈警告说，观察性横断面研究设计禁止任何因果解释，并提供了既不支持也不反对在 T1D 中使用 SGLTi 的有力证据。数据受限于使用 SGLTi 的参与者的比例很小（<3%）、可能的混杂因素以及注册表中收集的详细程度。¹重要的未知因素包括 SGLTi 使用与不良反应发生率之间的确切剂量、依从性和时间关系。有必要进一步研究检查与 T1D 中实际使用 SGLTi 相关的血糖结果和不利影响，特别是在实施风险缓解策略时。