Discussion
In this population-based study of patients with type 2 diabetes using metformin, use of SGLT2is was associated with a 25% reduced risk of the composite CV outcome when compared with non-SGLT2i use, which was driven by a reduced risk of IS. Use of SGLT2is was also associated with a clear 44% reduction in the risk of all-cause mortality, and an almost 50% reduction in the risk of severe renal disease. Because CKD was less frequent among initiators of SGLT2i, confounding could explain the observed decreased renal risk. However, we were able to observe identical effects irrespective of baseline CKD status, a noteworthy finding considering the large differences in background risk between these groups.
Overall, our findings are consistent with the CV and renal benefits of SGLT2is seen both in RCTs that apply strict eligibility criteria, and previous observational studies. Of note, by design, we explored CV benefits among individuals without previous CV disease, just as the renal benefits were explored among individuals free of severe renal disease at baseline. Therefore, it is important to highlight that our study was able to observe notable risk reductions among relatively young patients (ie, mean age of 60 years) at low CV risk. A recent meta-analysis based on six placebo-controlled RCTs found that SGLT2is were associated with a 10% reduction in the risk of major adverse CV events, and a 38% reduced risk of adverse kidney outcomes.18 A meta-analysis of cohort studies found SGLT2is to be associated with an 11% reduced risk of non-fatal IS versus DPP4is, a non-significant 7% reduction in the risk of MI, and 26% reduced risk of all-cause mortality.19 In that study, the magnitude of the risk reductions was increased to 21% for stroke and 37% for all-cause mortality when pooling OT estimates rather than those from their ITT analysis, in line with the lower HRs seen in our OT analyses. A recent large international cohort study using data from claims, medical records and national registries reported a 51% reduced risk of renal events among individuals initiating SGLT2is versus other glucose-lowering drugs.5 In another previous cohort study that compared SGLT2is with DPP4is using UK primary EHRs, Idris et al10 found the former to be associated with reduced risks of around 30% for all-cause mortality, 25% for CKD hospitalization in patients with no history of CV/renal disease, and 51% for CKD hospitalization in those with, or at high risk of, CV disease; no reduction in risk seen for MI. In the USA, Xie et al9 reported a 19% reduced risk of all-cause mortality among users of metformin taking SGLT2is versus sulfonylureas as add-on therapy. Interestingly, our study did not show any CV benefits of SGLT2i among individuals with CKD at baseline. Further research is needed to confirm this observation.
More than 460 million persons worldwide are currently living with diabetes, with type 2 accounting for around 90% of cases, and prevalence projected to increase.20 Furthermore, many individuals have pre-diabetes or undetected hyperglycemia at the level of diabetes. Renal complications are strongly associated with an increased risk of stroke, MI, heart failure and mortality in persons with type 2 diabetes,21–24 and a key goal for diabetes care is to reduce organ injury and obtain a life expectancy similar to persons without diabetes.25 It is therefore crucial to consider various treatment options based on the current evidence base, in addition to treatment costs.26 Among large RCTs of novel glucose-lowering agents on CV/renal outcomes, different treatments have shown divergent effects.27–31 Although no trial has performed a head-to-head comparison, several trials of DPP4is have failed to show any preventive renal effects.27 For GLP-1 analogs, a renal preventive effect of around 15%–20% has been shown, mainly due to effects on macroalbuminuria, but with no or little effect on renal function/ESRD.28 29 In contrast, SGLT2is have shown renal preventive effects, both in trials and real-world studies, and our present study suggests this effect is independent from baseline CKD status. As SGLT2is have also demonstrated a cardiorenal preventive effect in persons without diabetes, and in those with renal function <45 mL/min/1.73 m2 where the glucose-lowering effect is reduced, these beneficial effects must act via other mechanisms.32 One likely mechanism is through lowering of the intraglomerular pressure, which is essential for preventing further renal progression.32 Moreover, renal complications are associated with increased activation of the renin–angiotensin–aldosterone system, altered metabolism and increased inflammation—all detrimental for CV complications.33 34
Although SGLT2is show a preventive effect on renal complications, it is important to note that renal function still progresses significantly in treated patients receiving this drug treatment.5 Furthermore, a small number of patients do not tolerate SGLT2is particularly well. Owing to the increased risk of ESRD, CV disease and mortality in patients with impaired renal function, there is therefore a need for other complementary or alternative renal preventive treatments in type 2 diabetes care. Recently, finerenone, a drug that reduces mineralocorticoid receptor activation, has shown preventive effects on CKD progression and CV disease in patients with type 2 diabetes and established CKD, and represents one such additional treatment option.35 36
Strengths of our study include the use of a large sample size from a data source representative of the UK general population; hence, our results have good generalizability. Also, as chronic disease, such as diabetes, is largely managed in primary care, where all prescriptions are automatically recorded upon issue, medication use is likely to be well captured. We used different strategies of analysis—ITT, AT, and OT—as often used in clinical trials, as well as Fine and Gray models to account for the competing risk of death, and these showed our findings to be robust. Furthermore, we were able to use multivariable methods to adjust for potential confounding stemming from observed baseline imbalances. Some of them, such as previously mentioned differences in GLP-1 agonist use, could have favored SGLT2i initiators had we not used these methods. One study limitation is the reliance on recorded eGFR values to classify individuals by baseline CKD status and to identify severe renal events during follow-up. Although exposure to different glucose-lowering drugs could influence the frequency of renal function investigations, potentially biasing our results, virtually identical risk estimates were seen in the sensitivity analysis restricted to individuals whose renal function at baseline could be ascertained. One limitation is that renal function deterioration during follow-up could potentially affect SGLT2i use—either initiation or discontinuation of the drug for this very reason—and this would influence the results of the OT and AT analyses, where longitudinal change in drug use was considered. Also, it is not uncommon for people to discontinue life-extending medications in the last stages of life, and as we cannot exclude this as an explanation for the further reduced risk of mortality seen with current use of SGLT2is (as shown in the OT and AT analysis), caution is therefore needed when interpreting the time-dependent exposure mortality estimates. Also, causes of death are not systematically recorded in IMRD-UK, and accordingly, we could only identify a small number of CV deaths, which was insufficient for analysis as a separate endpoint. As noted earlier, the average follow-up was approximately 2 years, so evidence from this study regarding longer periods of use is scarce. We explored several endpoints using multiple strategies of analysis. While this might inflate type I error, consistency with previous evidence from trials and observational studies makes it unlikely that our results are explained by multiple testing. Finally, as we focused on endpoints within our CV composite outcome, we did not evaluate heart failure as explored in clinical trials.
In conclusion, our results indicate that among individuals with type 2 diabetes on metformin, use of SGLT2 is associated with significant CV, renal and survival benefits under normal conditions of use, confirming findings from RCTs on this topic. However, it is important to be aware of the level of unmet need that still exists in this patient population, as shown in the high incidence of CV and renal outcomes in this present study, especially among those with concurrent CKD.