5c.1 Background and Evidence Review
Acute kidney injury
The introduction of SGLT 2 inhibitors was accompanied by early concerns that their use may be linked to an increased risk of acute kidney injury (AKI) and volume depletion. This was largely driven by specific features of their mechanism of action, including an initial reduction in eGFR (which is, thereafter, followed by stabilisation of eGFR slope and improved kidney outcomes compared to placebo), induction of osmotic diuresis and natriuresis, alongside post-marketing reports of AKI events following their initiation. Of note, more than half of these reported AKI events occurred within the first 4 weeks of initiation. The US FDA dictated caution to health care professionals with regards to their use, especially in the context of other factors that may predispose to AKI such as CKD, heart failure and certain pharmacological agents such as angiotensin-converting enzyme inhibitors (ACEi), angiotensin-II receptor blockers (ARBs) and diuretics (34).
Since then, randomised placebo-controlled trials, including participants considered at high risk of AKI, have consistently demonstrated that SGLT 2 inhibition reduces, rather than increases, the risk of AKI. The recent meta-analysis by the RSG and SMART-C suggest reduced risk of AKI is now counted as a beneficial effect of SGLT 2 inhibition. After standardising analyses (wherever possible) to the MedDRA Preferred Term of “Acute Kidney Injury”, the meta-analysis demonstrated that SGLT 2 inhibition reduced the risk of AKI by 23% compared to placebo, independently of diabetes status, inclusive of people with heart failure and CKD, and unmodified by trial-level average eGFR (3). Such a reduction in risk translated to an estimated 4 fewer AKI events per 1000 person-years of SGLT 2 inhibition treatment among people with diabetes, or 5 fewer AKI events per 1000 person-years among people without diabetes (see Figure 5.1 above).
Initiation of SGLT 2 inhibitors is followed by a reduction in eGFR which is inherent to their mechanism of action (7, 8, 35, 38, 40, 41), accompanied by stabilisation of the eGFR slope within weeks (7, 35, 40, 41), and appears to be largely reversible upon discontinuation (36). In CREDENCE, the reported average eGFR decrease at 3 weeks was 3.72 ±0.25 mL/min/1.73m2 versus 0.55 ±0.25 mL/min/1.73m2 in the canagliflozin and placebo group, respectively, while in DAPA-CKD the eGFR decline at 2 weeks was 3.97 ±0.15 mL/min/1.73m2 in the dapagliflozin versus 0.82 ±0.15 mL/min/1.73m2 in the placebo group (7, 8). This initial eGFR decline, also referred to as ‘eGFR dip’, does not appear to have any clinical impact on AKI risk and does not appear to modify benefits or risks of treatment. In a post-hoc analysis of EMPA-REG OUTCOME, factors like diuretic use and worsening KDIGO CKD stage appear to predispose to a larger (>10%) eGFR dip 4 weeks after initiation of empagliflozin (42). Regardless, eGFR stabilised after 4 weeks and in the study the eGFR dip resolved upon discontinuation of study treatment. The treatment-mediated cardiovascular and renal benefits were not modified by the presence of a more pronounced eGFR dip, and eGFR remained stable from week 12 onward in all ‘eGFR dipping’ categories. In CREDENCE, the extent of eGFR drop did not affect the long-term change in eGFR slope, or the safety and tolerability of treatment (43). Likewise, an exploratory analysis of DAPA-CKD did not show any relationship between the magnitude of eGFR dip and kidney disease progression, hospitalisation for heart failure or cardiovascular death, or all-cause mortality, though there was evidence of attenuation of the chronic eGFR slope among dapagliflozin-treated participants in the highest category of eGFR dip (>10% reduction in eGFR) (44).
Hypovolaemia
CANVAS reported an increased rate of volume depletion events with canagliflozin compared to placebo (26.0 vs 18.5/1000 person years; p=0.009) (37), while CREDENCE found no significant excess in people with albuminuric diabetic kidney disease (28.4 vs 23.5/1000 person years) (7). More frequent episodes of volume depletion were reported with the use of dapagliflozin compared to placebo in DAPA-CKD (5.9 vs 4.2%; p=0.001) (8), but no significant excess was apparent in DECLARE-TIMI 58 (2.5 vs 2.4%) (21). Rates of adverse events consistent with hypovolaemia did not differ between the empagliflozin and placebo arms in EMPA- REG OUTCOME overall (35) or by eGFR categories (eGFR <60 vs ≥60 mL/min/1.73m2) (36), with similar findings for ertugliflozin in VERTIS CV (23). Serious or symptomatic dehydration events were comparable between empagliflozin (30/3304, 0.9%) and placebo (24/3305, 0.7%) arms in EMPA-KIDNEY (1).
In most of the large heart failure trials (2, 12, 13, 45, 46), hypovolaemia-related adverse event rates were similar between the treatment and placebo arms, with the vast majority of participants concurrently prescribed other diuretics. EMPEROR-PRESERVED found that hypotension (10.4% vs 8.6%) and symptomatic hypotension (6.6% vs 5.2%) were higher in the empagliflozin group than in the placebo group (14) , while the DELIVER trial showed similar rates of serious adverse events or adverse events leading to discontinuation suggestive of dehydration in the dapagliflozin and placebo arms (1.3% vs 1.0%) (2). In a subgroup analysis of data from DAPA- HF by diuretic dosage, volume depletion events were more common with dapagliflozin than with placebo in participants on the higher dose diuretics (47), however, there was no increase in renal adverse events.
Sotagliflozin, which also inhibits gut SGLT-1 and can cause diarrhoea, resulted in significantly more frequent volume-related adverse events than placebo in the SCORED trial in CKD and type 2 DM (24) (5.3 vs 4.0%; p=0.003), but with no significant excess reported in those with recent hospitalisation for worsening heart failure in SOLOIST-WHF (45) (9.4% vs 8.8%).
In a meta-analysis of the data from some of these large trials combined with a series of smaller trials, there was an increased risk of hypovolaemia-related trial adverse event reports with use of SGLT 2 inhibitors (OR=1.20, 95%CI 1.10-1.31) (48).
Potassium
Combining an SGLT 2 inhibitor with an ACEi or ARB does not have the same potential as dual renin- angiotensin system (RAS) blockade to cause hyperkalaemia (12, 13, 49). There were no meaningful differences in potassium between treatment groups on serial measurements in the CANVAS trial (49). There were also no reported significant differences in adverse events for hyperkalaemia between treatment groups in the large placebo-controlled CKD trials of SGLT 2 inhibitors (in which the majority of participants were treated with single RAS blockade). CREDENCE reported hyperkalaemia event rates of 29.7/1000 person-years among people allocated canagliflozin vs 36.9 events/1000 person-years for those allocated placebo (7). In DAPA-CKD, there were 6 (0.3%) serious adverse events of hyperkalaemia among those allocated dapagliflozin versus 12 (0.6%) among those allocated placebo (8). EMPA-KIDNEY reported serious hyperkalaemia among 92/3304 and 109/3305 participants treated with empagliflozin and placebo respectively (HR 0.83, 95%CI 0.63-1.09) (1).
Data from HFrEF populations are similarly reassuring, with no effect of SGLT 2 inhibitors on laboratory measurements of potassium or clinical events of hyperkalaemia overall, or among those co-prescribed mineralocorticoid receptor antagonists (MRA) (50, 51). These subanalyses from DAPA-HF and EMPEROR-REDUCED have generated hypotheses that SGLT 2 inhibition may even reduce the risk of severe hyperkalaemia among MRA users or lead to fewer discontinuations of MRA. Consistent with these findings, a large individual participant data meta-analysis of participants with diabetes from six SGLT 2 inhibitor trials, including four trials in enrolling participants with diabetes and high cardiovascular risk and two trials enrolling participants with CKD, demonstrated a significant 16% reduction in serious hyperkalaemia (95%CI 7% to 24%) among people treated with SGLT 2 inhibitors (52). The reduction in hyperkalaemia was present and of a similar magnitude across the spectrum of kidney function, including among people with eGFR <45 mL/min/1.73m2, and was not accompanied by an increased risk of hypokalaemia. Section 2 provides further details of data on hyperkalaemia with SGLT 2 inhibition among those with heart failure, including MRA users.
5c.2 Recommendations for Implementations
- We recommend that individuals initiated on an SGLT 2 inhibitor do not routinely require an early assessment of kidney function or serum potassium following initiation of treatment (Grade 1C).
- We suggest that if an individual has a kidney function assessment within the first few weeks post initiation of an SGLT 2 inhibitor, a decline in eGFR needs to be interpreted with caution and in the context of an expected drug effect to avoid unwarranted discontinuation of treatment (Grade 2B).
- We suggest that individuals on diuretics are counselled on the symptoms of hypovolaemia and advised to seek medical attention if they develop any such symptoms after starting SGLT 2 inhibition (Grade 2B).
- We suggest that clinicians consider an early clinical review and if appropriate a diuretic or antihypertensive dose reduction in individuals they consider at high risk of hypovolaemia (Grade 2C).
- We recommend that SGLT 2 inhibitors are temporarily withheld during acute illness (Grade 1C).
Rationale: SGLT 2 inhibitors have proven benefit in relation to reducing the rate of long-term decline in kidney function in certain groups of people with CKD. The means by which they provide this benefit may involve changes to intraglomerular pressure and reduction in hyperfiltration at an individual glomerulus level. This can result in a reduction in eGFR over the initial few weeks following initiation of SGLT 2 inhibitors, which is relatively small, largely reversible and should not usually be seen as an adverse effect of the drug.
None of the major studies have demonstrated an increased risk of AKI in people treated with SGLT 2 inhibitors, and it seems likely they have renal tubular protective effects that reduce risk of AKI. It is therefore important that early changes in eGFR that occur following initiation of SGLT 2 inhibitors do not routinely result in withdrawal of SGLT 2 inhibition when people are likely to gain significant benefit from them.
In addition, SGLT 2 inhibitors have a combined osmotic diuretic and natriuretic effect, so clinicians and the people treated with SGLT 2 inhibitors need to be aware of this effect in order to ensure that any risk of hypovolaemia is minimised.