Drugs used in diabetes
| Type | Recommendation |
|---|---|
| When to deprescribe | |
| CBR |
We suggest individualising HbA1c targets based on individual factors and preferences, aiming to avoid hypoglycaemia. In general, we support the suggested HbA1c targets:
|
| CBR |
We suggest that deprescribing decisions be made in collaboration with the individual and their diabetes care team, including specialist providers. For older people using diabetes medicines to manage glycaemic control in type 2 diabetes mellitus (T2DM), we suggest offering deprescribing:
|
| CBR |
In older people taking SGLT2 inhibitors or GLP-1 analogues for indications other than their glycaemic control benefits (e.g. cardiovascular and/or renal risk reduction), we suggest deprescribing be offered if these medicines are associated with adverse effects (e.g. potential muscle wasting which can exacerbate frailty), ensuring the benefit of discontinuation outweighs the risk and other management strategies are in place. We suggest that deprescribing decisions be made in collaboration with the individual and their diabetes care team, including specialist providers. |
| Ongoing treatment | |
| CBR |
We suggest continuing diabetes medicines used for glycaemic control in older people where the benefits generally outweigh the risks, including those who:
provided treatment aligns with the individual's goals and preferences, following informed consent. |
| How to deprescribe | |
| CBR |
We suggest discontinuing oral diabetes medicines without the need for tapering with the possibility of restarting the medicine if needed, provided this approach aligns with the individual's goals and preferences, following informed consent. Seek expert advice for the tapering of injectable diabetes medicines |
| CBR |
For people on combination therapy of diabetes medicines, we suggest:
For instance, prioritising short-acting insulins and sulfonylureas and last for other agents with additional cardiovascular risk reduction if considered appropriate to deprescribe. In people taking other medicines that impact blood glucose levels (e.g. centrally acting medicines, beta-blockers, thiazide diuretics, antipsychotics, corticosteroids, quinolones, ACE inhibitors), we suggest close monitoring of blood glucose levels for the first two weeks when deprescribing diabetes medicines. |
| Monitoring | |
| CBR |
We suggest monthly monitoring of the overall risk-benefit profile and lifestyle changes for at least three months, then every six months thereafter. However, this should be tailored based on individual factors such as their preferences, responses and tolerance to deprescribing. We suggest careful monitoring for signs and symptoms of hypoglycaemia and hyperglycaemia, as the presentation can often be different in older people and easily missed in older people. |
| CBR |
Self-monitoring of blood glucose For people who are already self-monitoring blood glucose, we suggest advising people to self-monitor random and fasting blood glucose concentrations at least once daily during tapering, and self-monitor for symptoms of hyperglycaemia (e.g. increased nocturia or thirst), as well as reporting to their healthcare provider if symptomatic or if their blood glucose concentration becomes elevated. HbA1c monitoring We suggest reviewing HbA1c levels once after approximately three months, and then twice a year for people who are stable and well-controlled. |
| GPS |
Continuous glucose monitoring (CGM) CGM should be considered and offered to people with diabetes to detect glucose fluctuations throughout the day, noting that CGM is not subsidised by the National Diabetes Services Scheme for individuals who do not have Type 1 diabetes at the time of writing. Where it is safe to do so, de-escalate blood glucose monitoring in line with patient preferences and goals of treatment to reduce the daily burden of disease monitoring (ungraded good practice statement). |
| GPS |
Healthcare providers should reinforce the benefits of optimal dietary intake and physical activity (ungraded good practice statement). |
CBR, consensus-based recommendation; GPS, good practice statement
The optimal intensity of glycaemic control for older people with type 2 diabetes is highly debated. Type 2 diabetes remission is possible and occurs more frequently in people over 75 years of age and people who had a substantial weight loss (over 15 kg) [189]. This could be due to reduced nutritional intake in older people coupled with age-related metabolic changes. As a result, adjustments in diabetes treatment are often necessary in this population to avoid overtreatment. Overly intensive glycaemic control (HbA1c < 7%) in older people has been associated with recurrent episodes of hypoglycaemia, which can have serious consequences, including an increased risk of cardiovascular and cerebrovascular events, cognitive decline, falls, and mortality [190-193]. Among diabetes medicines, insulin and sulfonylureas carry the highest risk of hypoglycaemia, with the risk heightened in older people due to the higher rate of renal or hepatic impairment, malnutrition, and low body weight which were found to be risk factors for hypoglycaemia [177, 194]. Hypoglycaemia in older people may have different presentation, with neurological symptoms such as dizziness, visual disturbances, agitation, confusion, or behavioural changes being more prominent than autonomic symptoms. In people with dementia, these neurological symptoms can be misinterpreted as dementia-related symptoms [195]. Furthermore, symptoms of hypoglycaemia are less specific with increasing age and asymptomatic hypoglycaemia is also common in older people, further complicating its management. While hypoglycaemia poses significant risks, undertreatment of diabetes can increase the likelihood of both microvascular and macrovascular complications.
Refer to the narrative evidence summary, the GRADE Summary of Findings table in the guidelines, and the Technical Report for a complete presentation of the deprescribing evidence based on the GRADE framework (including other factors considered in developing the recommendations).
Glycaemic control
Below we summarised the findings from key trials that contributed to the guidelines for glycaemic control in older people. Findings from the studies suggest that deintensifying glycaemic control in older people or those with longstanding diabetes may be unlikely to worsen microvascular outcomes in the short term. However, before de-intensifying treatment, it is essential to review medicines that can affect glycaemic control in people with diabetes. Some medicines are known to cause hyperglycaemia, including thiazide diuretics, atypical antipsychotics (particularly olanzapine), and corticosteroids. Conversely, alcohol is associated with hypoglycaemia in addition to other medicines such as salicylates, quinolones, beta-blockers, and ACE inhibitors.
For individuals who are already self-monitoring their blood glucose, checking random or fasting blood glucose levels after deprescribing, such as at least once daily, can help track glycaemic control [196]. However, the frequency of monitoring should be individualised based on the person's medication regimen, clinical stability, and preferences. Where safe and appropriate, blood glucose monitoring can be de-escalated in line with the individual’s goals and preferences to reduce the daily burden of disease management [197]. When encouraging self-monitoring, it may be helpful to provide examples of common symptoms of hyperglycaemia (e.g. increased thirst or nocturia), as many people may not recognise that these non-specific symptoms could indicate elevated blood glucose. Continuous glucose monitoring (CGM) should be considered and offered as an alternative to identify trends in blood glucose levels and reduce the need for frequent fingerstick testing. Additionally, HbA1c levels reflect glycaemic control over the preceding six to eight weeks [198]. It may be appropriate to test HbA1c once around three months after deprescribing, and then every six months for individuals who are stable and well-controlled.
The United Kingdom Prospective Diabetes Study (UKPDS) was a pivotal RCT that compared intensive glycaemic control (fasting plasma glucose, FPG < 6 mmol/L) with conventional control (FPG < 15 mmol/L) in 5,102 patients with newly diagnosed type 2 diabetes. Participants had a median age of 54 years (IQR 48-60) and were followed for over 10 years while receiving intervention with diet alone (conventional), metformin, sulfonylureas, or insulin (intensive) [199]. The study found that intensive glycaemic control with metformin, sulfonylureas, or insulin reduced the risk of microvascular complications but had no significant effect on macrovascular disease (approaching statistical significance, p = 0.052) during the trial period.
A follow-up conducted 10 years after the UKPDS RCT concluded demonstrated long-term benefits, with a reduced rate of microvascular disease, myocardial infarction, and mortality in those who had received intensive treatment [200]. This glycaemic "legacy effect" suggests the potential long-term benefits of early intensive glycaemic control.
Participants in the UKPDS were approximately 10 years younger at baseline compared to those in the ADVANCE (Action in Diabetes and Vascular Disease: Preterax and Diamicron Modified Release Controlled Evaluation) and the ACCORD (Action to Control Cardiovascular Risk in Diabetes trials) (mean age = 66 and 62 years respectively) [201, 202]. The mean diabetes duration was 8 years in ADVANCE and 10 years (median) in ACCORD. In contrast to UKPDS, these two studies found that intensive glycaemic control was associated with an increased risk of mortality. These differences in age and diabetes duration have important implications when adjusting treatment intensity, particularly in the context of deprescribing.
Similarly, the 2009 Veterans Affairs Diabetes Trial (VADT) included 1,791 military veterans (mean age = 60 years) with type 2 diabetes who were nonresponsive to at least one oral diabetes medicine at the maximum dose and/or daily insulin injections, with nonresponse defined as a central HbA1c ≥ 7.5% or local HbA1c ≥ 8.3% [203]. The mean diabetes duration was 11.5 years. This study found no significant differences between the intensive and standard glycaemic control groups in major cardiovascular events, cardiovascular mortality, all-cause mortality, or microvascular complications, except for a reduced progression of albuminuria in the intensive group.
Table 8. Possible HbA1c target (adapted from ElSayed et al., 2023 [204])
| Possible HbA1c target | Populations |
|---|---|
| < 7.0 - 7.5% (53-58 mmol/mol) | Robust older people (two or fewer co-existing chronic conditions, intact cognitive and functional status) |
| < 8.0 % (64 mmol/mol) | Older people with complex/intermediate health status (three or more co-existing chronic conditions requiring medicines/lifestyle interventions, two or more instrumental ADL impairments, or mild-to-moderate cognitive impairment) |
| Avoid specifying strict HbA1c targets | Older people with complex or poor health (moderate-to-severe cognitive impairment, two or more impairments in activities of daily living, chronic illnesses with significant symptoms/impairment of functional status, limited life expectancy) as symptom management and quality of life may be more relevant than HbA1c targets. |
International guidelines provide general glycaemic goals for older people (see Table 8) but emphasise the importance of individualising these goals based on unique characteristics through shared decision-making to address individual needs and preferences [205]. Other important factors that should be considered include frailty, diabetes duration and presence of cardiovascular diseases as discussed above, in addition to hypoglycaemia awareness, history of severe hypoglycaemia, diabetes-related distress, and concerns such as fear of hypoglycaemia [206]. For example, a robust person aged 70 years with no established cardiovascular diseases may aim for an HbA1c target of less than 7%. If a person had a 15-year history of diabetes or an established cardiovascular disease, a less intensive target of less than 7.5%, or even below 8% in the presence of additional comorbidities or hypoglycaemia risk, may be considered.
Cardiovascular and/or renal benefits
As discussed, many RCTs of diabetes medicines were conducted prior to the introduction of newer agents (SGLT2 inhibitors and GLP-1 receptor agonists). These newer agents have demonstrated cardiovascular or renal benefits independent of their benefits in glycaemic control. As such, less intensive glycaemic targets, combined with the careful selection of these medicines, may offer a safer and more favourable benefit-risk profile for certain people.
Below we summarised the findings from key trials for SGLT2 inhibitors and GLP-1 receptor agonists in the context of cardiovascular and/or renal benefits.
SGLT2 inhibitors
A systematic review and meta-analysis that included 35 RCTs that assessed the cardiovascular effects of SGLT2 reported that SGLT2 inhibitors significantly reduce the incidence of mortality, major adverse cardiac events, non-fatal myocardial infarction and heart failure in patients with Type 2 diabetes [207]. Among the studies included in the review, the 2015 EMPA-REG OUTCOME trial (Empagliflozin, Cardiovascular Outcomes, and Mortality in Type 2 Diabetes) was a prominent study [208]. This randomised, double-blind, placebo-controlled trial included 7,020 patients (mean age = 63 years) with Type 2 diabetes and at high risk for cardiovascular events. Patients were included if they had an HbA1c of between 7% - 9% (for those who had not received glucose-lowering agents for at least 12 weeks before randomisation) or 7% - 10% (for those who had received stable glucose-lowering therapy for at least 12 weeks before randomisation). Participants who were randomised to 10 mg or 25 mg of empagliflozin (an SGLT2 inhibitor), in addition to standard care, had significantly lower cardiovascular mortality, all-cause mortality, and hospitalisation for heart failure compared to placebo. However, there were no significant between-group differences in the risk of nonfatal myocardial infarction or nonfatal stroke and hospitalisation for unstable angina. In terms of adverse events, there was a significant increase in genital infection among patients who received 10 mg or 25 mg empagliflozin compared to placebo but no significant difference between the two groups in other adverse events including hypoglycaemia, acute renal failure, diabetic ketoacidosis, thromboembolic events, bone fracture, and events consistent with volume depletion.
The 2022 EMPA-KIDNEY trial (Study of Heart and Kidney Protection with Empagliflozin) randomised 6,609 patients (mean age = 64 years) with chronic kidney disease (CKD), a wide range of GFRs, levels of albuminuria, and causes of CKD to receive empagliflozin (10 mg once daily) or placebo [209]. The study included patients with or without diabetes. There was a significantly lower risk of progression of kidney disease or death from cardiovascular causes in the group who received empagliflozin. Results were consistent among patients with or without diabetes. The group who received empagliflozin also had a significantly lower risk of all-cause hospitalisation. In the subsequent 2024 follow-up study involving 4,891 (74%) surviving patients who consented, findings suggest that empagliflozin may have residual cardiorenal benefits for up to 12 months after it was discontinued [210].
The 2017 CANVAS (Canagliflozin and Cardiovascular and Renal Events in Type 2 Diabetes) trial assessed the effects of the SGLT2 inhibitor canagliflozin on cardiovascular, renal, and safety outcomes in 10,142 participants with Type 2 diabetes and high cardiovascular risk [211]. The 2019 CREDENCE (Canagliflozin and Renal Events in Diabetes with Established Nephropathy Clinical Evaluation) trial assessed the effects of the SGLT2 inhibitor canagliflozin on renal outcomes in 4,401 participants with type 2 diabetes and albuminuric chronic kidney disease [212]. Both trials (both mean age = 63 years) reported that canagliflozin significantly reduced the risk of cardiovascular death, myocardial infarction, stroke, and heart failure hospitalisation. These studies consolidated previous findings of the renal and cardiovascular protection of SGLT2 inhibitors.
More recently, a 2025 post hoc analysis utilised the data from the CANVAS and CREDENCE trials to assess the efficacy and safety of SGLT2 inhibitors in people with frailty [213]. From the pooled, individual participant-level data analysis, frailty (defined as Frailty Index > 0.25) was present in 56% of all participants in the two trials (n = 10,142). This post hoc analysis reported that the benefits of canagliflozin in improving cardiovascular and mortality outcomes were observed in people with type 2 diabetes irrespective of their frailty status, and canagliflozin did not further increase the risk of adverse events except for osmotic diuresis (where the opposite was reported, i.e. osmotic diuresis was less common in frail participants compared with non-frail). However, it is important to note that all participants had a significantly higher risk of adverse events (including fracture, volume depletion, osmotic diuretic, amputation, diabetic ketoacidosis, and genital infection) with canagliflozin use. In these trials, only 0.2% (25/14,543) of all participants were underweight with a BMI < 18.5 whereas the majority of the participants were overweight (4608/14,543; 32%) or obese (8375/14,543; 58%).
An individualised approach is needed to optimise therapy for people with frailty. Evidence suggests the benefits of the SGLT2 and GLP-1 agonists in people who are underweight, anorexic, and malnourished are lacking as they are often underrepresented or excluded in clinical trials [214]. It is possible that some individuals may not be able to tolerate these newer therapies due to the associated risk of causing further weight loss, dehydration, and hypotension that may further exacerbate frailty.
GLP-1 receptor agonists
A 2025 meta-analysis that included 11 RCTs that assessed the cardiovascular and renal effects of GLP-1 receptor agonists reported that GLP-1 receptor agonists led to a significant reduction in clinically important kidney events, kidney failure, and cardiovascular events [215]. Among the studies included in the meta-analysis, the Semaglutide Effects on Cardiovascular Outcomes in People with Overweight or Obesity (SELECT) and the Evaluate Renal Function with Semaglutide Once Weekly (FLOW) trials were two recent trials conducted in 2023 and 2024 respectively.
In the 2023 SELECT trial, 17,604 patients (mean age = 62 years) with no diabetes who had pre-existing cardiovascular disease, overweight or obese (body-mass index of ≥ 27) were randomised to receive once-weekly semaglutide (a GLP-1 receptor agonist) or placebo and followed up for a mean of 40 months [216]. Participants who had received semaglutide had a significantly lower risk of primary composite outcome (cardiovascular mortality, nonfatal myocardial infarction, or nonfatal stroke). However, the semaglutide group had experienced a significantly higher rate of adverse events leading to permanent discontinuation.
In the FLOW trial, 3,533 patients (mean age = 67 years) with type 2 diabetes and CKD who were at very high risk for kidney disease progression, kidney failure, cardiovascular events, or death (according to the Kidney Disease: Improving Global Outcomes risk calculators) were included [217]. Compared to placebo, The group who had received semaglutide had a significantly lower risk of clinically important kidney outcomes and death from cardiovascular causes in patients with type 2 diabetes and chronic kidney disease.
We identified five studies (four cohort studies and one before-and-after study) related to deprescribing diabetes medicines from the systematic review and meta-analysis [169, 218-221].
Overall, there is limited evidence suggesting significant benefit or harm associated with the discontinuation of diabetes medicines. Most studies were cohort studies, and although one of these studies showed that deprescribing was associated with a reduction in hypoglycaemic episodes, the certainty of this evidence is very low and insufficient to support evidence-based recommendations.
Key study characteristics and results
A narrative summary of each study is provided below, highlighting key characteristics and main findings.
Sjöblom 2008 included 32 older Swedish nursing home residents with type 2 diabetes and intensive (HbA1c ≤ 6%) glycaemic control in a prospective cohort study [221]. Diabetes medicines remained deprescribed in 24 (75%) participants three months after the initial dose reduction or withdrawal. At six months, the mean HbA1c in the deprescribing group was lower, 5.8 ± 1.1% compared to 6.6 ± 1.4% in the continuation group. However, continuation group participants had a higher baseline HbA1c compared to the intervention group (7.1 ± 1.6% vs 5.2 ± 0.4%). Four participants in the deprescribing group reported hyperglycaemia (OR 21.0, 95% CI 1.09, 403.01) and subsequently dropped out of the study. Their plasma glucose levels were 14.6, 16.6, 17.4, and 18.3 mmol/L respectively.
Hui 2019 conducted a retrospective study that included older people with type 2 diabetes who have a prescription for diabetes medicines other than metformin and either two most recent consecutive HbA1c ≤ 6.5% or ≤ 7.0% but had either visited the emergency department or were hospitalised for hypoglycaemia in two years prior. The study reported that the discontinuation of diabetes medicines significantly reduced the risk of hypoglycaemia (OR 0.46, 95% CI 0.24, 0.90) and all-cause mortality (OR 0.40, 95% CI 0.24, 0.69) [218]. Additionally, there was no significant difference in the proportion of patients who experienced hyperglycaemia between the deprescribing group and the continuation group (OR 0.43, 95% CI 0.13, 1.43).
Niznik 2022 conducted a retrospective cohort study that included veteran nursing home residents with advanced dementia or limited life expectancy with HbA1c ≤ 7.5%. Type 1 and type 2 diabetes could not be distinguished using administrative data. The study reported that deintensifying diabetes medicines (dose reduction or discontinuation) was not associated with all-cause emergency department visits, hospitalisation, or death in weighted analyses for 60 days after deintensification [219].
Silverii 2020 conducted a retrospective cohort study that included outpatients with type 2 diabetes aged over 75 years. Within the cohort in which deprescribing of diabetes medicines was performed (n=46), there was a reduction in the rate of severe hypoglycaemia six months following deprescribing (none versus five cases in the prior six months) while mean HbA1c increased significantly from 6.4 ± 2.6% to 7.0 ± 3.3% (p < 0.05) [220].
Linsky 2022 targeted diabetes medicines and PPIs by mailing patient-centred educational brochures to veterans two weeks prior to their scheduled primary care appointments. Targeted veterans were either taking a PPI for at least 90 consecutive days or were at an increased hypoglycaemia risk (diabetes diagnosis with a prescription for insulin or sulfonylurea, most recent HbA1c < 7%, and either aged 65 or over, had renal insufficiency, or cognitive impairment). Medicine appropriateness was not determined from administrative data as the goal of the study was to promote discussion of deprescribing. The study reported that the intervention group was more likely to have the target medicine discontinued or reduced (14% versus 4%, p = 0.009) and have discussions with their healthcare providers about the target medicine (12% versus 1%, p = 0.001) [169].
There was a lack of description on the method of deprescribing in most studies and no direct evidence indicates tapering or abrupt discontinuation was associated with the greatest benefits and harms.
Four studies reported important/critical outcomes of very low certainty. In the study by Hui 2019, withdrawal schedules were likely individualised following a pharmacist-led assessment (n=2740) [218]. In the study by Sjöblom 2008, diabetes medicines were discontinued abruptly except for insulin over 20 units/day for which the dose was reduced by half (n=32) [221]. Methods were not described in the other two studies (n=2128) [219, 220].
The other study by Linsky 2022 did not report important or critical outcomes associated with deprescribing and the method of deprescribing was not described (n=348) [169].