In this retrospective study, we have linked data from the Swedish NDR with records from other national healthcare registries and performed propensity score-weighted analyses to compare large cohorts of adults with type 2 diabetes on insulin therapies who self-manage their glucose levels using either isCGM or BGM.

The first observation of note is that isCGM users with type 2 diabetes on insulin therapy achieve comparable and sustained reductions in HbA1c after initiating isCGM, irrespective of treatment with either intensive multiple daily injections or basal insulin-only regimens. Compared with control groups using BGM, both the T2D-MDI cohort and the T2D-B cohort demonstrated HbA1c reductions of 3.7 mmol/mol (0.34%) and 3.5 mmol/mol (0.32%), respectively, 6 months after starting isCGM, and this was maintained for 24 months in both cohorts (Fig. 1). These data support earlier studies showing that use of isCGM in people with type 2 diabetes is associated with lower HbA1c, whether they are treated with multiple daily injections [5, 6], basal insulin only [10, 11] or any glucose-lowering therapies [34]. Notably, in absolute terms, the improvements in HbA1c for isCGM users recorded in our study are also almost identical to those observed in previous RCTs in which the efficacy of real-time CGM vs BGM on glucose management was assessed in individuals with type 2 diabetes treated with multiple daily injections [4] or basal insulin only [9]. Moreover, for isCGM users with suboptimal glycaemic management and baseline HbA1c ≥58.5 mmol/mol (≥7.5%), reductions in HbA1c were higher than for the overall population of isCGM users with type 2 diabetes on basal insulin or multiple daily injections (Fig. 1 and ESM Table 3). For isCGM users in the T2D-MDI and T2D-B cohorts with HbA1c ≥70 mmol/mol (≥8.6%), reductions in HbA1c from baseline were even greater (ESM Table 3), with reductions in HbA1c of 14.3 mmol/mol (1.31%) for both cohorts after 6 months, which were sustained at 24 months. Again, the reductions in HbA1c were comparable between the T2D-B and T2D-MDI cohorts for individuals with suboptimal glycaemic management.

The second clinically important observation relates to our analysis of the hospital admissions data (Tables 2, 3 and 4), which showed that the entire group using isCGM (T2D-MDI and T2D-B) had significantly lower risk of hospitalisation for severe hypoglycaemia (−57%), stroke (−44%), heart failure (−37%) and acute non-fatal myocardial infarction (−33%) relative to BGM control participants (Table 2). Risk of hospitalisation for any reason was also significantly reduced in the whole isCGM user group (−29%) (Table 2), and in both the T2D-MDI and the T2D-B subgroups using isCGM, compared with the BGM control groups (−16% and −24%, respectively) (Tables 3 and 4). These outcomes are comparable with similar reductions in hospital admission rates for any cause reported in two studies of people with type 2 diabetes, one after initiating isCGM [35] and the other using any CGM method [36].

However, the multiple daily injection and basal insulin-only regimens showed different risk profiles for hospitalisation for known complications of diabetes. First, the data demonstrated a reduction in hospitalisation for severe hypoglycaemia only in the T2D-MDI cohort (−49%, p=0.034) but not in the T2D-B cohort (Tables 3 and 4). In the RELIEF study, using the French national health claims database, hospitalisation for hypoglycaemia over the 2 year study period was reduced both for people with type 2 diabetes on any therapy (43% fewer admissions), of whom 85% were on intensive insulin therapy [37], but also for the smaller subgroup with type 2 diabetes on basal insulin only (44% fewer admissions) [38]. This difference in results may reflect a more conservative approach to using the different insulin regimens in type 2 diabetes in Sweden compared with France, as the ratio of T2D-MDI and T2D-B therapies was more equal in our study, and the hospital event rates for hypoglycaemia in the T2D-B control cohort (0.25 per 100 person-years, Table 4) were lower than those reported in the RELIEF study cohort for the 12 months prior to starting isCGM (0.73% of study population). Nonetheless, the comparable reductions in admission rates for severe hypoglycaemia in people with type 2 diabetes on multiple daily injections seen in both studies clearly suggest a preventive effect of isCGM on the acknowledged increased risks of severe hypoglycaemia for people with type 2 diabetes on intensive insulin therapy. In accordance with these results, another study reported a 29% reduction in hospital admissions for hypoglycaemia for people with type 2 diabetes on prandial insulin after starting isCGM [39]. To our knowledge, no studies have reported on admissions for severe hypoglycaemia in type 2 diabetes after commencing use of other CGM devices.

Second, regarding hospitalisation for major cardiovascular events, such as stroke, heart failure and non-fatal acute myocardial infarction, for adults with type 2 diabetes, we found reduced risk of admission for stroke and non-fatal acute myocardial infarction for adults in the T2D-MDI cohort (−46% and −25%, respectively) and a 37% reduced risk of admission for heart failure in the T2D-B cohort. While we believe our study is the first to report the impact of using isCGM or other CGM technologies on alleviating the risks for hospitalisation for these macrovascular complications of type 2 diabetes, the underlying reasons for this favourable finding, and for the different outcomes in the intensive and non-intensive insulin treatment groups, remain to be elucidated. Major risk factors for CVD, such as lipids, BP and renal function, were adjusted for in the analysis and should not have had a bearing on the results. Although it cannot be completely ruled out, it seems unlikely that these risk reductions can be solely explained by the improvement in glycaemic management after initiating isCGM, given the rather modest difference in HbA1c between the isCGM and BGM groups and the relatively short follow-up period. As the risk of major cardiovascular events for people with type 2 diabetes is associated with hypoglycaemia, including nocturnal and severe hypoglycaemia [40, 41], it may be proposed that the reduced risk of severe hypoglycaemia in adults with type 2 diabetes in Sweden who are using isCGM is associated with a reduced occurrence of major cardiovascular events. Other possible reasons for this reduced occurrence include an increase in time in range and reduced glucose variability [19,20,21,22]. However, this possibility requires further research and detailed analyses of CGM metrics, which were not available in this study.

It is important to acknowledge that our study has limitations. As use of isCGM is only one of a number of variables in our retrospective cohort study, it is possible that other factors may have influenced the observed outcomes. For example, the NDR data do not record whether the isCGM user received device training or education from their healthcare professionals around the index date, which may have resulted in improved diabetes self-care behaviours that are not controlled for in this retrospective analysis. Furthermore, the study does not contain data on duration of insulin use, which could have had an effect on HbA1c reduction. However, the groups were well balanced after the PS-IPTW for all measured variables, with weighted standardised differences below 0.10, which indicates a negligible treatment group difference and is not considered to be confounding, or below 0.20, which indicates only small differences between the covariates, such that any residual confounding is unlikely to affect the outcomes in a meaningful way.

It must also be acknowledged that the validity of the outcomes from an investigation using a diabetes registry and a wider patient registry relies on the quality of interpretation and application of the variables within each of the registries. For example, our control group had an older mean age at baseline compared with the isCGM intervention group; however, the weighted standard difference between these two groups was below 0.2, which is indicative of only a small difference between these groups. Diabetes duration and HbA1c were similar between the two groups, and the baseline prevalence of cardiovascular complications was similar or even higher in the isCGM users than in the BGM control participants, with weighted standard differences for all these covariates being lower than 0.1. Strengths of our study include the large population of people with type 2 diabetes on insulin therapies included, and the fact that the NDR, SPDR and NPR have greater than 90% coverage of all adults with diabetes in Sweden. Another strength is that specialised nurses provide education to people with type 2 diabetes before starting isCGM in Sweden, which ensures that use of the system is appropriate.

As the clinical type of diabetes that a person has is stored at the person level in the NDR and is updated to be the last recorded observation, there should be no error in the variable containing clinical diabetes type. Thus, interpretation of our study outcomes is not confounded by errors in classification of diabetes type. This means that selection bias should not be a factor. An important strength of our analysis is the inclusion of control groups, identified from within the NDR, that match the baseline characteristics of our study cohorts across a wide range of variables, but who were not incident users of isCGM or other CGM devices between June 2017 and August 2022. This has allowed us to control for selection bias and to mitigate the impact of other factors on glycaemic management over the study period, independent of isCGM.

This real-world retrospective cohort study of adults with type 2 diabetes in Sweden, using analysis of data from the NDR linked with comprehensive prescribing information from the SPDR, has shown that adults with type 2 diabetes on intensive insulin therapy with multiple daily injections and adults with type 2 diabetes treated with basal insulin only achieve significant and comparable reductions in HbA1c 6 months after being prescribed isCGM, and these reductions persist for at least 24 months. Furthermore, both in adults with type 2 diabetes treated with multiple daily injections and adults with type 2 diabetes treated with basal insulin, the reductions in HbA1c are positively correlated with the HbA1c at baseline.

In addition, we have been able to demonstrate a significantly lower risk of hospitalisation for severe hypoglycaemia or for major cardiovascular events, or hospitalisation for any reason, for adults with type 2 diabetes following initiation of use of the isCGM system compared with control participants who are using BGM. These outcomes have direct consequences for optimising health-related outcomes for people with type 2 diabetes, and have implications for the long-term cost-effectiveness of providing access to isCGM for people with type 2 diabetes on insulin therapies.