Values, preferences, and treatment burden for initiation of GLP-1 receptor agonists, SGLT-2 inhibitors, tirzepatide, and finerenone in adults with type 2 diabetes: systematic review

Abstract

Objectives

To summarise existing evidence regarding the values, preferences, and perceived treatment burden of adults with type 2 diabetes when considering initiation of glucagon-like peptide-1 receptor agonists (GLP-1 RA), sodium-glucose cotransporter-2 inhibitors (SGLT-2i), finerenone, or tirzepatide.

Design

Systematic review.

Data sources

Medline, Embase, Scopus, Web of Science, and Cochrane Central Register of Controlled Trials were searched from database inception to 7 January 2025 (updating a previous systematic review that captured literature up to 23 May 2020).

Eligibility criteria

Quantitative or qualitative studies were included if they reported patient reported values, preferences, or treatment burden among adults with type 2 diabetes and compared initiation of GLP-1 RA, SGLT-2i, tirzepatide, or finerenone with each other or with other glucose-lowering treatments.

Data extraction and synthesis

Paired reviewers independently screened studies, extracted data, and assessed risk of bias. Certainty of evidence was evaluated using GRADE. Findings were synthesised narratively.

Results

Thirty four studies (15 656 patients) were included, 17 of which were new since the previous 2021 review (8360 patients). Of the newly included studies, 10 evaluated GLP-1 RA, seven evaluated SGLT-2i, and one evaluated tirzepatide; no studies evaluated finerenone (because several studies compared multiple drug classes, these categories are not mutually exclusive). Across comparisons, patients consistently reported glycaemic efficacy, cardiovascular risk reduction, and simplicity of administration as their priorities. Oral once-daily regimens and once-weekly injectable treatments with easy-to-use devices were generally preferred. Weight loss was valued but was typically secondary to glycaemic and cardiovascular benefits. SGLT-2i and GLP-1 RA were often preferred over dipeptidyl peptidase 4-inhibitors (DPP-4i), thiazolidinediones, and insulins, but no consistent preference emerged for SGLT-2i over GLP-1 RA. Tirzepatide was favoured over semaglutide in a single study, largely due to greater haemoglobin A_1c reduction and lower hypoglycaemia risk. Certainty of evidence for all comparisons was low or very low certainty.

Conclusion

Adults with type 2 diabetes prioritise treatment effectiveness and convenience among glucose-lowering treatments. Current evidence does not support a clear preference between SGLT-2i and GLP-1 RA, and data for finerenone are absent. These findings underscore the importance of shared decision making informed by individual patient priorities.

PROSPERO registration

CRD42023429170.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• New diabetes medications have shifted the focus into cardiovascular risk reduction, while highlighting the need of individualised management

• Patients tend to prefer oral medications or once-weekly injectables, and often resist complex injection devices

• High quality evidence of how patients weigh the benefits, harms, and burden of treatment is limited

WHAT THIS STUDY ADDS

• No high certainty evidence suggests a consistent patient preference for any specific drug class over another

• Adults with type 2 diabetes prioritise improved glycaemic control, medication ease of use, reliable drug delivery, reduced dosing frequency, and fewer adverse events

• Convenient, reliable, and easy-to-use injection devices were highly valued among adults with type 2 diabetes

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE, OR POLICY

• The findings are based mostly on low certainty evidence

• Future studies need to minimise methodological bias by use of validated elicitation instruments and report attributes presented to patients more transparently, based on the best evidence

• Inclusion of more diverse populations and analyses of subgroups across multiple sociodemographic factors could improve the representativeness and applicability of findings to real-world settings

Introduction

Patient centred care is a core principle of contemporary type 2 diabetes.^1–3 Treatment decisions increasingly require balancing multiple outcomes, such as glycaemic control, cardiovascular and kidney risk reduction, and adverse effects, against patients' values and their experience of treatment burden and convenience.^4–9 Understanding patient values and preferences is therefore essential for shared decision making and for the development of trustworthy clinical practice guidelines.

Our previous systematic review of 17 studies (7292 patients), published in 2021, reported sparse and very low certainty evidence regarding patient values and preferences for newer glucose-lowering treatments and largely focused on delivery characteristics of glucagon-like peptide-1 receptor agonists (GLP-1 RA). No studies investigated preferences for sodium-glucose cotransporter-2 inhibitors (SGLT-2i), and few studies considered outcomes, beyond glycaemic control, that were important to the patient.¹⁰ As a result, identified evidence was insufficient to support nuanced comparisons between drug classes or to comprehensively inform guideline recommendations.

Since then, the therapeutic landscape for type 2 diabetes has changed substantially. Newer treatments, including the dual GIP/GLP-1 receptor agonist tirzepatide and the non-steroidal mineralocorticoid receptor antagonist finerenone, have shown important benefits for cardiovascular and kidney outcomes. Tirzepatide has also provided substantial weight reduction. Additionally, guideline recommendations have shifted away from a glucose-centric approach and towards prioritising long term risk reduction. Whether patients value these broader outcome benefits over treatment burden, adverse effects, and mode of administration is unclear.

We, therefore, conducted an updated systematic review to synthesise evidence on the values, preferences, and perceived treatment burden of adults with type 2 diabetes when considering initiation of SGLT-2i, GLP-1 RA, finerenone, or tirzepatide. This review updates and expands on our previous summary of available evidence¹⁰ and provides patient centred context to inform contemporary clinical decision making and guideline development.

This systematic review directly informed a BMJ Rapid Recommendation living practice guideline addressing medications for adults with type 2 diabetes, where evidence regarding benefits and harms, perceived patient values and preferences, and prognosis were accounted for in producing trustworthy recommendations regarding the four prioritised treatments (GLP-1 RA, SGLT-2i, tirzepatide, and finerenone).¹¹ BMJ Rapid Recommendations represents a collaborative effort between the MAGIC Evidence Ecosystem Foundation (http://www.magicevidence.org) and The BMJ to produce trustworthy recommendations in response to practice changing evidence. The parallel BMJ Rapid Recommendations guideline panel comprised of patient partners, methodologists, general practitioners, internists, endocrinologists, cardiologists, and nephrologists. These contributors helped to define the research question and the scope of this review. The panel included people free of financial competing interests, and any potential intellectual competing interests were minimised.

Methods

Study design and protocol registration

We conducted a systematic review in accordance with the 2020 Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) and 2021 Preferred Reporting Items for Reporting Literature Searches in Systematic Reviews (PRISMA-S) guidelines.^12,13 The protocol was registered in the International Prospective Register of Systematic Reviews (PROSPERO: CRD42023429170). This review updates and expands on a previous systematic review¹⁰ by extending the search to January 2025 and including tirzepatide and finerenone.

Eligibility criteria

We included quantitative or qualitative studies that published on patient reported values, preferences, or treatment burden among adults (>18 years) with type 2 diabetes. Studies had to compare GLP-1 RA, SGLT-2i, tirzepatide, or finerenone with one another or with other glucose lowering treatments. We excluded cost-effectiveness studies, studies without patient reported data, single arm studies, studies restricted to participants with prespecified treatment preferences, and randomised trials that assessed preferences based solely on prior treatment experience.

Data sources and search strategy

With assistance from an experienced information specialist, we searched English language studies in Ovid Medline, Embase, Scopus, Web of Science, and the Cochrane Central Register of Controlled Trials (CENTRAL) from database inception to 7 January 2025. A previously validated filter for studies on patient values and preferences was applied.¹⁴ Reference lists of included studies were screened to identify additional eligible studies. Full search strategies are provided in the online supplemental file 1.

Study selection

After excluding duplicate studies and completing a pilot exercise to ensure adequate inter-rater reliability, paired reviewers screened titles and abstracts using standardised forms, followed by full text reviews of potentially eligible studies. Discrepancies were resolved by consensus or discussion with a senior reviewer (RR-G). Study selection was performed using Distiller Systematic Review Software (Evidence Partners, Ottawa, Canada).

Data extraction

Following the pilot that was done to ensure adequate inter-rater agreement, paired reviewers used a standardised form to independently extract data from eligible studies. Extracted data included study design, participant characteristics, preference elicitation methods, treatment comparisons, attributes presented to participants, and reported preferences. Discrepancies were resolved by consensus or discussion with a senior reviewer (RR-G).

Outcome definition

In line with the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) working group, we considered values and preferences to include the relative importance that patients assign to potential benefits, harms, costs, inconvenience, and treatment burdens when choosing between management options.^15,16 When reported, we extracted preferences related to efficacy, cardiovascular and kidney outcomes, adverse effects, dosing frequency, route of administration, and device characteristics.

Risk of bias and certainty of evidence

Paired reviewers independently adjudicated the risk of bias of individual studies using a tool produced by the GRADE working group for values and preferences studies. This tool addressed participant selection, measurement methods, data completeness, and analysis.¹⁷ Disagreements were resolved by consensus or discussion with a senior reviewer (RR-G).

Certainty of evidence for each comparison was assessed using the GRADE approach, considering study design, risk of bias, inconsistency, indirectness, imprecision, and other methodological considerations.^17–19 Overall certainty of evidence was rated as being either very low, low, moderate, or high.

Data synthesis

Given heterogeneity in study designs, preference elicitation methods, and reported outcomes, we synthesised findings narratively. We did not perform a meta-analysis. We refer to medications primarily by drug class (eg, GLP-1 RA and SGLT-2i) and specify individual drugs when studies made direct comparisons.

Patient and public involvement

This study was a systematic review and did not involve the direct recruitment or participation of patients. Therefore, there were no study participants to whom the results could be directly disseminated. The patient and public involvement took place as part of the development of the guideline, which was subsequently informed by the findings of this review. The results of this review will be disseminated through the usual academic dissemination channels, including publication in a peer-reviewed journal. Where applicable, the guideline will also communicate the relevant findings to patient and public communities. Also, two patient partners informed the first iteration of the linked living BMJ Rapid Recommendation practice guideline for which this review provides supporting evidence, and which informed the scope of this work.¹¹

Results

Study selection and characteristics

The search identified 22 246 records, of which 34 studies (15 656 patients) met our inclusion criteria.^20–53 Seventeen new studies (8360 patients) were identified since the previous review (figure 1).^37–53

Figure 1. Study selection flow chart.

Study characteristics

Our included studies were published between 1 November 2010 and 18 November 2024 and enrolled adults with type 2 diabetes from Asia, Europe, North America, and South America. Sample sizes ranged 58-3327 participants. The mean age of participants ranged 47-65 years. Most studies included populations mixed with those who had no previous injection history and those with a history of receiving injectable diabetes medications. Preference elicitation methods included online surveys (n=3),^20,39,40 discrete choice experiments (n=19),^{22,24,25,27,29,30,33,35,37,38,42–47,50,51,53} time trade-off analyses (n=5),^{21,28,32,34,41} willingness-to-pay studies (n=3),^22,46,52 and cross-over designs (n=4).^26,36,48,49 Of the 34 studies, the evidence was heavily concentrated on GLP-1 RA, which were evaluated on 31 studies (15 749 participants).^{20–37,39–42,44–47,49–53} Nine studies (5695 participants) included SGLT-2i,^{38,42–44,46–48,50,51} and only one (620 participants) evaluated tirzepatide.⁴⁵ These categories are not mutually exclusive, as several studies compared multiple therapeutic classes within the same elicitation method (eg, four studies compared SGLT-2i v GLP-1 RA). No studies assessed patient preferences for finerenone (tables 1 and 2). Twenty eight (82%) of the 34 included studies reported pharmaceutical industry funding.^21–47,53 A deeper analysis of findings revealed a consistent pattern whereby manufacturer-sponsored studies frequently reported preferences favouring the sponsor's product. Details on funding and extended versions of tables 1 and 2, including all studies of the original review, may be found on the online supplemental material.

Table 1. Demographic and study characteristics.

Author, year	Country	N	History of injection	Mean age, years (SD)	Female,%	Ethnic group (%)	Method	Drugs evaluated
Igarashi et al, 202037	Japan	500	No history	61.2 (9.3)	6.2	ND	DCE	Semaglutide QD, dulaglutide QW, liraglutide QD
Ozdemir et al, 202038	Singapore	147	Mixed	58.8 (12.3)	47	Chinese: 66.0; other: 34.0	DCE	SGLT-2i, DPP-4i
Boye et al, 2021a§39	USA; UK	600 (USA: 300; UK: 300)	Mixed	64 (11.3)	41.8	White: 93.3; black: 3.6; Asian: 2.3; American Indian or Alaska native 0.5; other: 0.7	Online survey	Dulaglutide QW, semaglutide QD
Boye et al, 2021b*40	USA; UK	504 (US: 253; UK: 251)	Previous history	USA 59.2 (13.1); UK 50.9 (15.5)	USA 44.7; UK 28.7	ND	Online survey	GLP-1 RA, insulin
Matza et al, 2021§41	UK	201	Mixed	58.7 (8.3)	47.3	White 80.1; black: 5.0; Asian 8.5; mixed ethnicity: 4.5; other: 2.0	TTO	Simple oral medication QD, dulaglutide QW, semaglutide QD, semaglutide QW
Banjara et al, 202242	USA	176	Mixed	60.4 (13.7)	54.6	White: 86.9; Hispanic: 6.8; Black: 3.4; Asian: 1.7; Alaska native 0.6; other: 0.6	DCE	SGLT-2i, GLP-1 RA
Costa-Gil et al, 202243	Argentina; Mexico	390	Mixed	57.5 (12.3)	53.1	ND	DCE	DPP-4i QD, SGLT-2i QD
Shubrook et al, 202244	USA	500	Mixed	65.5 (ND)	42.6	ND	DCE	GLP-1 RA, empagliflozin QD sitagliptin QD, thiazolidinedione QD
Gelhorn et al, 2023*§45	USA; UK	620 (USA: 301; UK: 319)	No history	USA 60.7 (12.7); UK 58.9 (12.5)	USA 50.8; UK 50.5	USA: white 89.7; black 8.3; Asian 3.7; Alaska native 3.7; native Hawaiian 0.3; UK: English/Welsh/Scottish/Northern Irish/British 92.8; white and black Caribbean 6.0; white and Asian 2.2; Irish 1.3; Indian 1.3; black British 0.9; white and black African 0.9; Caribbean 0.6; Gypsy or Irish traveller 0.3; Pakistani 0.3; Bangladeshi 0.3; Chinese 0.3; African 0.3	DCE	Tirzepatide QW, semaglutide QW
Panchal et al, 202346	USA	58	Mixed	47.5 (12.0)	63.8	White: 77.6; other: 13.8; missing: 8.6	DCE	SGLT-2i, GLP-1 RA, DPP-4i, insulin, premixed insulin, sulphonylureas, thiazolidinedione
Savarese et al, 202347	Sweden	553	Mixed	<55: 15% of population, 55-64: 36%, ≥65: 49%	60.4	ND	DCE	Empagliflozin QD, semaglutide QD, sitagliptin QD
Shields et al, 202348	UK	457	No history	61.8 (9.5)	26	White: 94.8; Asian: 2.6; black: 0.4; other: 1.3; not stated: 0.9	Crossover	Pioglitazone QD, sitagliptin QD, canagliflozin QD
Wettergreen et al, 202349	USA	60	Mixed	61 (ND)	50	White: 65.0; Hispanic: 18.3; Native American: 3.3; black: 3.3; Asian: 1.7; other: 3.3; preferred not to answer: 5.0	Crossover	Dulaglutide QW, semaglutide QW, exenatide QW
Liu et al, 202350	China	3327	Mixed	ND	50.8 (n=1691)	ND	DCE	Drug attributes corresponding: insulin, GLP-1 RA, SGLT-2i, sulphonylureas, nateglinide, meglitinide, biguanides, thiazolidinedione, alpha-glucosidase inhibitor, DPP-4i
Vaidya et al, 2024a†51	India	87	Mixed	57.2 (11.4)	44.8 (n=39)	ND	DCE	Glimepiride, pioglitazone, tenegliptin, dapagliflozin, liraglutide, insulin glargine
Vaidya et al, 2024b†52	India	87	Mixed	57.2 (11.4)	44.8 (n=39)	ND	WTP	Semaglutide QD, insulin QW
San José et al, 202453	Spain	180	Mixed‡	63.35 (11.4)	36.7 (n=66)	ND	DCE, WTP	GLP-1 RA injection devices

^*Demographic characteristics reported by country (USA and UK).

^†Same set of participants.

^‡Previously untreated for GLP-1 RA.

^§Percentages may not sum up to 100% because categories were not mutually exclusive across all studies. Data were extracted as reported in the original articles.

DCE, discrete choice experiment; DPP-4i, dipeptidyl peptidase-4 inhibitors; GLP-1 RA, glucagon-like peptide-1 receptor agonists; ND, not described; QD, once daily administration; QW, once weekly administration; SD, standard deviation; SGLT-2i, sodium-glucose co-transporter-2 inhibitors; TTO, time trade-off; WTP, willingness to pay.

Table 2. Drug evidence profiles presented to participants.

Author, year	Change in HbA1c %	Adverse effects, %	Weight change	Hypoglycaemia, %	Dosing frequency	Type of delivery system	Macrovascular and microvascular outcomes
Igarashi et al, 202037	MF: −2, H1: −1.8 H2: −1.6, LF: −1.4	Nausea (without vomiting) MF: 9, H1: 6 H2: 4, LF: 2	MF: −2.4 kg, H1: −1.5 kg H2: −0.6 kg, LF: 0.3 kg	ND	Semaglutide, QD; dulaglutide, QW; liraglutide, QD	Semaglutide, oral; dulaglutide, single-use injection; liraglutide, multiple-use injection	ND
Ozdemir et al, 202038	ND	UGTI SGLT-2i: 5, DPP-4i: 2	SGLT-2i: −2 kg, DPP-4i: none	ND	ND	ND	Admission to hospital for heart failure, SGLT-2i: 1%, DPP-4i: 3%; Risk of death SGLT-2i: 1%, DPP-4i: 3%
Banjara et al, 202242	Level 1: 10 Level 2: 50 Level 3: 90	Chances of GI side effects Level 1: 1, Level 2: 15, Level 3: 30; Changes in genital infection Level 1: 0, Level 2: 8, Level 3: 16	ND	ND	Level 1: QD, Level 2: QD, Level 3: twice daily, Level 4: QW	Level 1: oral, Level 2: injectable, Level 3: injectable, Level 4: injectable	Major adverse cardiovascular events, Level 1: no reduction, Level 2: 20%, Level 3: 40%
Costa Gil et al, 202243	DPP-4i: −0.9 SGLT-2i: −1.2	Risk of genital infections DPP-4i: 0 SGLT-2i: 1	DPP-4i: 0 kg SGLT-2i: 2.4 kg	DPP-4i: low risk SGLT-2i: low risk	DPP-4i: QD SGLT-2i: QD	DPP-4i: oral SGLT-2i: oral	Admission to hospital for heart failure, DPP-4i: 0% risk reduction, SGLT-2i: 27; kidney disease, DPP-4i: 0 risk reduction, SGLT-2i: 47
Shubrook et al, 202244	Semaglutide QD : −1.50, semaglutide QW: −1.60, dulaglutide: −1.20, empagliflozin: −0.90, sitagliptin: −0.80	ND	Semaglutide QD: −9.5 lb, semaglutide QW: −9 lb; dulaglutide: −4 lb, empagliflozin: −8 lb; sitagliptin: −1.5 lb	Semaglutide QD: 3, semaglutide QW: 3, dulaglutide: 3, empagliflozin: 2.3, sitagliptin: 1.6	Semaglutide: QD, semaglutide: QW, dulaglutide: QW, empagliflozin: QD, sitagliptin: QD	Semaglutide: oral or injectable, dulaglutide: injectable pen, empagliflozin: oral, sitagliptin: oral	Semaglutide QD: −21, semaglutide QW: −26, dulaglutide: −12, empagliflozin: −14, sitagliptin: no effect, thiazolidinediones: no effect
Gelhorn et al, 202345*	Semaglutide: 66, TIR5: 74, TIR10: 82, TIR15: 87	Nausea (mild to moderate) Semaglutide: 17.9, TIR5: 17.4, TIR10: 19.2, TIR15: 22.1	Semaglutide: 6.2 kg, TIR5: 7.8 kg, TIR10:10.3 kg, TIR15: 12.4 kg	Semaglutide: 4/1000, TIR5: 6/1000, TIR10: 2/1000, TIR15: 17/1000	Semaglutide: QW, TIR5: QW, TIR10: QW, TIR15: QW	Semaglutide: multi-use pen, all tirzepatide: single-use pen,	ND
Panchal et al, 202346	Level 1: −0.5, Level 2: −1, Level 3: −1.5	GI effects (nausea, vomiting, diarrhoea) L1: 0, L2: 25, L3: 50	ND	ND	ND	Oral or injected	Level 1: heart failure RRR 0%, cardiovascular events RRR 0%; Level 2: 25%, 15%; Level 3: 50%, 30%
Savarese et al, 202347	Empagliflozin: −0.9, semaglutide: −1.4, sitagliptin: −0.8	Risk increase of UTI, vomiting and nausea	Empagliflozin: −8.4 lb, semaglutide: −9.3 lb, sitagliptin: −1.5 lb	ND	Empagliflozin: QD, semaglutide: QD, sitagliptin: QD	Empagliflozin: oral, semaglutide: oral, sitagliptin: oral	Cardiovascular death, empagliflozin: substantial evidence for risk reduction; semaglutide: some evidence; sitagliptin: no change
Liu et al, 202350	Insulin: −2.5; GLP-1 RA, SGLT-2i, insulin: −1.5; sulphonylureas, nateglinide, biguanides: −1; thiazolidinediones, AGI, DPP-4i: −0.5	GI effects DPP-4i, SGLT-2i: 0; meglitinide: 10; AGI: 20; GLP-1 RA: 40	GLP-1 RA, SGLT-2i: −2 kg, AGI, DPP-4i: 0 kg; sulphonylureas, nateglinide, thiazolidinediones: 1.5 kg; insulin: 3 kg	Thiazolidinediones, AGI: 0; nateglinide: 5; sulphonylureas: 15; insulin: 30	ND	Insulin, most GLP-1 RA: injection, others: pill	Cardiovascular benefit, GLP-1 RA, SGLT-2i: yes; others: no
Vaidya et al, 202451	A: 1.2-1.8; B: 1-1.5; C: 0.8-1; D: 0.8-1.5; E: 1.1-2; F: 1.5-2	A: allergies; B: oedema, heart failure, atypical fractures; C: respiratory infections, GI effects, headache; D: infection (genital and UTI); E: GI; F: injection reaction	Multiple weight gain and loss options	A, B, F: high risk; C, D, E: low risk	A, B: once or twice daily; C, D, E, F: once daily	A, B, C, D: oral tablet; E, F: subcutaneous injection	B, D, E: risk reduction of MACE (myocardial infarction, stroke, heart failure), admission to hospital, death; reduces progression of CKD
Vaidya et al, 2024b52	Insulin: −1.33 reduction; semaglutide: −1.8 decrease (after six months)	Insulin: mild weight gain, low risk of injection site reactions, hypersensitivity; semaglutide: nausea, diarrhoea, pancreatitis	Insulin: mild weight gain; semaglutide 14 mg: −3.7 kg	Insulin: might occur; semaglutide: rare	Insulin: once weekly; semaglutide: once daily	Insulin: injected; semaglutide: oral	Semaglutide: risk reduction of atherosclerotic cardiovascular disease, stage 3 CKD, or heart failure with reduced ejection
San José et al, 202453	ND	ND	ND	ND	Daily, weekly	No preparation; simple preparation; complex preparation	ND

^*Data for USA participants shown in table, data for UK shown in online supplemental file 1.

AGI, alpha-glucosidase inhibitors; CKD, chronic kidney disease; DPP-4i, dipeptidyl peptidase-4 inhibitors; GI, gastrointestinal; H1, hypothetical profile 1; H2, hypothetical profile 2; LF, least favourable; MF, most favourable; ND, no data; QD, once daily administration; QW, once weekly administration; RRR, relative risk reduction; SGLT-2i, sodium-glucose co-transporter-2 inhibitors; TIR 5, tirzepatide 5 mg; TIR 10, tirzepatide 10 mg; TIR 15, tirzepatide 15 mg; UGTI, urinary and genital tract infections; UTI, urinary tract infection.

Risk of bias and certainty of evidence

Most studies were judged as having high risk of bias. The most common concerns were primarily due to the use of non-validated preference elicitation tools, insufficient reporting of drug profiles presented to participants, high attrition, or limited reporting of analytical methods.^{20–27,29,31,34,35,38–40,42,44,45,47–49,51–53} Only a few studies used validated preference elicitation tools and drug profiles that were clearly based on the best available evidence, raising concerns about the validity of preference estimates (figure 2).^{28,30,32,33,36,37,41,43,47,50} Overall, certainty of evidence was rated as low or very low across all comparisons due to concerns regarding risk of bias, indirectness and imprecision, and the absence of confidence intervals and other measures of uncertainty (table 3). An extended version of table 3, including detailed explanations supporting the judgements made, can be found on the online supplemental material.

Figure 2. Risk of bias assessment.

Table 3. GRADE assessment of the certainty of evidence by comparison.

Comparison	No of studies	Study design	Risk of bias	Inconsistency	Indirectness	Imprecision	Other considerations	Effect	Certainty
Tirzepatide v GLP-1 RA (semaglutide)	1	Observational study	Very serious	Not serious	Very serious	Serious	None	Tirzepatide was preferred over semaglutide	Very low
GLP-1 RA (liraglutide, dulaglutide, lixisenatide, semaglutide, exenatide) v insulin (rapid acting, short acting, intermediate acting, long acting, premixed insulin products)	6	Observational studies	Very serious	Not serious	Very serious	Very serious	None	GLP-1 RA was preferred over insulin in four studies and insulin was preferred in one study; the other study evaluated injection devices: devices that were easier to use overall and provided confidence of correct and complete dose delivery were preferred	Very low
SGLT-2i (canagliflozin, empagliflozin, ertugliflozin, dapagliflozin) v thiazolidinedione (pioglitazone)	4	Observational studies	Very serious	Not serious	Very serious	Very serious	None	SGLT-2i was preferred over thiazolidinedione	Very low
GLP-1 RA (liraglutide, lixisenatide, semaglutide, dulaglutide, exenatide) v DPP-4i (alogliptin, linagliptin, sitagliptin, saxagliptin, vidagliptin, teneligliptin)	7	Observational studies	Very serious	Very serious	Very serious	Very serious	None	GLP-1 RA was preferred in two studies and DPP-4i in four studies; the remaining study did not show a preference for either class	Very low
SGLT-2i (empagliflozin, canagliflozin, dapagliflozin, ertugliflozin) v DPP-4i (sitagliptin, alogliptin, linagliptin, saxagliptin)	7	Randomised & Observational Studies	Serious	Not serious	Very serious	Serious	None	Drug profiles similar to SGLT-2i were preferred over DPP-4i	Very low
GLP-1 RA (liraglutide, lixisenatide, semaglutide, dulaglutide) v TZD (pioglitazone, rosiglitazone)	4	Observational studies	Very serious	Not serious	Very serious	Very serious	None	GLP-1 RA was preferred over thiazolidinedione. In two studies, participants were willing to pay more for GLP-1 RA compared with other drugs; in one study participants were willing to pay more specifically for liraglutide	Very low
SGLT-2i (empagliflozin, canagliflozin, ertugliflozin, dapagliflozin) v insulin (rapid acting, short acting, intermediate acting, long acting, premixed insulin products)	2	Observational study	Serious risk	Not serious	Very serious	Very serious	None	Participants preferred and were willing to pay the most for SGLT-2i profiles with proven cardiovascular benefits and none was willing to pay for various insulin products	Very low
Dulaglutide (QW injectable) v semaglutide (QD oral, QD injectable)	7	Observational studies	Very serious	Not serious	Very serious	Serious	None	After presenting information regarding treatment process and injection device of QD oral semaglutide compared with QW injectable dulaglutide, semaglutide was preferred in three studies; when comparing the treatment process and injection devices between QW injectable semaglutide and QW injectable dulaglutide, dulaglutide was preferred in four out of five studies; one study compared treatment process, injection devices, and clinical data for QW injectable semaglutide with QW injectable dulaglutide and semaglutide was preferred	Very low
Liraglutide v semaglutide	1	Observational studies	Not serious	Not serious	Very serious	Serious	None	QD oral semaglutide was preferred over QW injectable liraglutide when treatment processes and clinical data are presented	Very low
SGLT-2i (empagliflozin, canagliflozin, ertugliflozin, dapagliflozin) v GLP-1 RA (dulaglutide, semaglutide, liraglutide, lixisenatide)	4	Observational studies	Very serious	Serious	Serious	Serious	None	In two of four studies, empagliflozin was preferred; in another study, oral semaglutide was preferred, followed by injectable semaglutide; in a study that compared drug class in general, oral once daily treatment was preferred; one study preferred dapagliflozin over liraglutide	Very low
GLP-1 RA injection devices	5	Observational studies	Serious	Not serious	Very serious	Serious	None	As administration requirements for GLP-1 RA injection devices increase, preferences decrease; weekly injection devices were strongly preferred over daily; preferred injection devices were those that were easier to use and that provided confidence of correct and complete dose delivery	Very low

Detailed explanations supporting the judgements made in the table can be found on the online supplemental material.

GLP-1 RA, glucagon-like peptide-1 receptor agonist(s); GRADE, Grading of Recommendations Assessment, Development, and Evaluation; QD, once daily administration; QW, once weekly administration; SGLT-2i, sodium-glucose co-transporter-2 inhibitors; TZD, thiazolidinediones.

Preferences by drug class comparison

Table 3 presents GRADE assessments capturing comparison level findings and related certainty of evidence ratings (online supplemental material). Tables 2 and 4 summarise drug evidence profiles and attributes presented to participants in the included studies.

Table 4. Drug preferences and main attributes leading to preference among included studies.

Author, year	Drug preference (as measured)	Unit of measurement	Scale	Attributes (attribute weight)
Igarashi et al, 202037	Oral semaglutide, 14 mg: 64%; oral semaglutide, 7 mg: 62.4%; injectable dulaglutide, 0.75 mg: 13.6%; injectable liraglutide, 0.9 mg: 11%	Relative importance (attribute importance)	Continuous numerical value / 0-100%	Mode of administration (daily oral tablet): 1.72 (49.1%); risk of nausea: 1.08 (30.8%); weight change: 0.40 (11.3%); HbA1c change: 0.31 (8.8%)
Ozdemir et al, 2020†38	SGLT-2i: 88%; DPP-4i: 12%	Relative importance	0-100%	Class 1 Mortality risk: 54%, risk of admission to hospital from heart failure: 20%, risk of UGTI: 15%, years of medication effectiveness (2, 4, or 6 years): 10%, weight loss: 7%, cost: 0% Class 2 Risk of UGTI: 30%, cost: 29%, mortality risk and years of medication effectiveness (2, 4, or 6 years): 14%, risk of admission to hospital from heart failure: 11%, weight loss: 2%
Boye et al, 2021a‡39	Semaglutide QD: 76.5%; dulaglutide QW: 23.5%	Patient preferences according to drug choice	0-100%	Semaglutide Oral administration: 47.9%; already taking oral medications and can add another: 24.2%; concerned about using injectable medication: 11.3%; daily administration: 9.8%; other: 3.5%; fits better into schedule: 3.3% Dulaglutide Weekly administration: 78%; already taking oral medication and do not want another oral medication: 14.2%; injected administration: 3.5%; fits better into schedule: 2.1%; other: 1.4%; burdensome to carry pills: 0.7%
Boye et al, 2021a§39	Semaglutide QD: 52.2%; dulaglutide QW: 47.5%	Patient preferences according to drug choice	0-100%	Switched oral to injectable (n=153) Timing/steps of taking the medication: 58.8%; convenience: 52.9%; frequency the medication is taken: 40.5%; how the medication fits better with my schedule: 40.5%; injection device: 33.3%; concerns about properly administering the medication: 15%; experience with injectable: 5.2%; other: 4.6%
Boye et al, 2021b40	ND	Attribute importance	0-100%	Confidence in administering the correct dose: 59.5%; ease of selecting the correct dose of medication: 53.2%; overall ease of use of the injection device: 47.4%; frequency of injections (eg, daily v weekly): 44.2%; convenience/portable device: 37.7%; ease of storage of injection device and medication: 35.1%; ease of preparation of injection device for injection: 34.4%
Matza et al, 2021¶41	Daily oral semaglutide: 53.2%; weekly injectable dulaglutide: 46.8%	Utility	0-1; 0=dead, 1=full health	Simple oral medication 0.89 (95% CI 0.870 to 0.909); semaglutide oral 0.880 (0.860 to 0.9); dulaglutide injection 0.878 (0.857 to 0.898); semaglutide injection 0.859 (0.836 to 0.881)
Banjara et al, 202242	SGLT-2i or GLP-1 RA as an oral, once daily treatment	Relative importance	Continuous numerical value	Out-of-pocket cost: 2.2; reaching target HbA1c in six months: 1.4; mode of administrations (tablet, once a day): 0.8; percentage RR of major adverse cardiovascular events: 0.7; gastrointestinal side effects: 0.6; genital infection: 0.4
Costa Gil et al, 202243	SGLT-2i 88.2%; DPP-4i: 11.8%	Ranking of attribute importance	1-5	Most important attribute was HbA1c reduction of 67.4% for DPP-4i; HbA1c reduction of 43.6% for SGLT-2i; Least important attribute was genital infection risk of 37% for DPP-4i and body weight reduction of 37%; genital infection risk of 68.9% for SGLT-2i
Shubrook et al, 2022**44	Oral semaglutide (91.9%) v dulaglutide (8.1%); oral semaglutide (70.1%) v empagliflozin (29.9); oral semaglutide (83.8%) v sitagliptin (16.2%); oral semaglutide (86.8%) v thiazolidinedione (13.2%); injectable semaglutide (89.3%) v dulaglutide (10.7%); injectable semaglutide (55.7%) v empagliflozin (44.3%); injectable semaglutide (64.5%) v sitagliptin (35.5%); injectable semaglutide (68.3%) v thiazolidinedione (31.7%)	Relative importance	0-100%	Change in weight: T: 29.2%; E: 41%; N: 25.4%; mode and frequency of administration: T: 35.3%; E (injection QW): 27.7%; N: (oral QD) 39.5%; cardiovascular RR: T: 19.1%; E: 22.1%; N: 16.7%; change in HbA1c: T: 6.5%; E: 6.96%; N: 6.14%; risk of hypoglycaemic events: T: 9.9%; E: 0.2%; N:12.3%
Gelhorn et al, 2023††45	USA: tirzepatide, 10 mg: 51.4%; tirzepatide, 15 mg: 24.3%; tirzepatide, 5 mg: 19.9% semaglutide, 1 mg: 4.4%	Relative importance	0 to 100%	HbA1c reduction: 26.3%; delivery system (single use prefilled pen ready for injection): 20.9%; weight change: 18.3%; hypoglycaemia: 18%; frequency of nausea: 16.4%
Panchal et al, 202346	Participants were willing to pay more for empagliflozin compared with all other drugs	Utility coefficient (SE); willingness to pay for a 1-unit level change of attribute	0 to no limit	Reduction in HbA1c: 0.037; reduction in risk of heart failure: 0.014 ($7.14); reduction in the risk of cardiovascular events: 0.011 ($5.82); out-of-pocket cost: −0.0019; chance of gastrointestinal side effects: −0.013 (-$7.06); dosing flexibility: −0.065; route of administration: −0.15
Savarese et al, 202347	Empagliflozin 25 mg: 41%; sitagliptin 100 mg: 31%; oral semaglutide 14 mg: 11%	Relative importance	0-100%	Risk of UGTI: 19%; weight reduction: 15%; fasting requirements and food restriction: 15%; risk of vomiting: 14%; RR of cardiovascular death by more than a third: 12%; risk of nausea: 11%; HbA1c reduction: 8%; can be taken at the same time as other oral medications: 6%
Shields et al, 2023‡‡48	Canagliflozin: 37.2%; sitagliptin: 33.3%; pioglitazone: 23.6%; no preference: 5.9%	Ranking of attribute importance	0-100%	Feeling better (the drug had a positive effect on health or wellbeing): canagliflozin: 63%; sitagliptin: 48%; pioglitazone: 39%; lack of side effects: canagliflozin: 27%; sitagliptin: 42%; pioglitazone: 50%
Shields et al, 2023§§48	Canagliflozin: 38.3%; sitagliptin: 34.6%; pioglitazone: 25.2%	Ranking of attribute importance according to drug choice	1-3 (one most preferred and three least preferred)	HbA1c: canagliflozin: 25.6%; sitagliptin: 70%; pioglitazone: 76.9%; weight: canagliflozin: 18.6%; sitagliptin: 2.5%; pioglitazone: 2.6%; both HbA1c and weight: canagliflozin: 37.2%; sitagliptin: 17.5%; pioglitazone 7.7%
Wettergreen et al, 202349	Injectable dulaglutide: 74.6%; injectable semaglutide: 13.6%; exenatide XR: 11.9%	Preference for GLP-1 RA injection devices	0-100%	Ease of use: dulaglutide: 74.6%; semaglutide: 13.6%; exenatide XR: 11.9%; Confidence dose delivery: dulaglutide: 72.9%; semaglutide: 13.6%; exenatide XR: 13.6% Exterior design and styling: dulaglutide: 60.7%; semaglutide: 26.8%; exenatide XR: 12.5% Size and portability: dulaglutide: 28.1%; semaglutide: 61.4%; exenatide XR: 10.5% Ease of dose setting: dulaglutide: 66.1%; semaglutide: 23.2%; exenatide XR: 10.7% Requires most effort to inject: exenatide XR: 53.4%; semaglutide: 39.7%; dulaglutide: 6.9%
Liu et al, 202350	Patients were willing to pay $33.6 USD for a medication with an efficacy of 2.5% point reduction in HbA1c over one with efficacy of 0.5% point reduction	Utility coefficient/WTP	0 to no limit	HbA1c reduction 2.5%: 1.378 (SE 0.082); hypoglycaemia risk 0%: 1.271 (0.078); cardiovascular benefit: 1.172 (0.060); GI adverse effect 0%: 1.138 (0.078); weight change −2 kg: 0.098 (0.067); pill: 0.913 (0.056)
Vaidya et al, 2024a51	Glimepiride (51.75); dapagliflozin (22.9); tenegliptin (17.2); liraglutide (4.6); pioglitazone (2.3); glargine (1.5)	Quantitative patient preference	1-100%	Effect on weight (29%); route of administration (23%); additional benefits offered by the drug (18%)
Vaidya et al, 2024b52	Insulin QW: Rs157.25 (±112.60); oral semaglutide once daily: Rs9.35 (±5.66)	Willingness to pay	0 to no limit	ND
San José et al, 202453	Patients were willing to pay €83.25 to change a dose with a “somewhat complex preparation” to one with “no preparation required”	Utility coefficient/willingness to pay	0 to no limit	Preparation of the dose not needed: 3.152; simple preparation of the dose: 1.677; once weekly: 1.310; at any time or daily: 0.841

An extended version of table 4 that includes all studies of the original review can be found on the online supplemental material.online supplemental file 1

^*Before being fed with HbA_1c and weight results.

^†A 2-class latent class model was used to estimate the maximum acceptable risk of UGTI for various treatment benefits. Class one ranked all-cause mortality risk as the most important attribute. Class two ranked UGTI risk as the most important attribute (30%) and cost as a close second (29%).

^‡Initial reasons for choosing semaglutide.

^§Preferences after learning about the medication administration process through a video (post-video reasons for preference of total population are not reported).

^¶Health state vignettes eliciting treatment processes with GLP-1 RA regarding dosing, administration, and requirements of medications were presented through a video to elicit patient preferences through a questionnaire.

^**Relative preference weights reported separately by subgroups (T: total population; E: injection experienced population; N: no history of injection).

^††Data for UK shown on online supplemental material.

^‡‡Before being fed with HbA_1c and weight results.

^§§Preferences after being provided with HbA_1c and weight results.

^¶¶Shown for 125 participants who changed preference on second ranking after being fed with HbA_1c and weight results.

UGTI, urinary and genital tract infections.DPP-4i, dipeptidyl peptidase-4 inhibitors; E, injection experienced population; GI, gastrointestinal; GLP-1 RA, glucagon-like peptide-1 receptor agonists; N, no history of injection; ND, not described; QD, once daily administration; QW, once weekly administration; RR, risk reduction; Rs, Indian rupee; SE, standard error; SGLT-2i, sodium-glucose co-transporter-2 inhibitors; T, total population.

Across studies, patients consistently prioritised glycaemic efficacy, cardiovascular risk reduction, and simplicity of treatment administration when choosing between glucose lowering treatments. These attributes outweighed concerns about gastrointestinal adverse effects or weight change in most comparisons. Oral once daily regimens and once weekly injectable treatments were generally preferred to treatments requiring multiple daily injections, and simple devices were prioritised.

GLP-1 RA versus DPP-4i

Seven observational studies compared GLP-1 RA with DPP-4i.^{20,23,26,44,46,47,51} Most studies had results that favoured DPP-4i; however, preferences were inconsistent: results from two studies showed a preference for GLP-1 RA,^23,44 while four favoured DPP-4i.^20,26,44,47 In the remaining study, neither DPP-4i nor GLP-1 RA were selected as the preferred option.⁴⁶ Attributes that led to a preference for DPP-4i were route of administration and reduced frequency of adverse effects, whereas for GLP-1 RA preference, the attributes were greater efficacy and weight loss. Certainty of evidence was very low, downgraded for very serious risk of bias, inconsistency, indirectness, and imprecision.

GLP-1 RA versus insulin

Six observational studies assessed preferences between GLP-1 RA and insulin.^{22,33,34,40,50,52} In four studies, participants preferred GLP-1 RA, primarily due to perceived benefits in glycaemic control, risk of adverse cardiovascular and heart failure, weight change, and easier administration (delivery system and dosing frequency).^22,33,34,40 In one study where most participants preferred GLP-1 RA, a subset of individuals preferred insulin primarily due to a lower risk of gastrointestinal adverse effects and pancreatitis.³³ One study showed a preference for insulin, although the results did not specify which attributes influenced patient preferences;⁵² and one study focused on injection device characteristics rather than drug preference.⁵⁰ Certainty of evidence was very low, downgraded for serious risk of bias, indirectness, and imprecision.

GLP-1 RA versus tirzepatide

One observational study compared tirzepatide with semaglutide.⁴⁵ Participants preferred tirzepatide across three tested doses of tirzepatide, driven primarily by a greater HbA_1c reduction, an easier delivery system, and a lower risk of hypoglycaemia. Certainty of evidence was very low, downgraded for very serious risk of bias and indirectness, and for serious imprecision.

GLP-1 RA versus SGLT-2i

Four observational studies compared SGLT-2i with GLP-1 RA.^42,44,46,47 Preferences were inconsistent. Results from two studies favoured SGLT-2i (particularly empagliflozin and dapagliflozin), driven primarily by an HbA_1c reduction and a reduced risk of heart failure and cardiovascular events.^46,47 In one study, oral or injectable semaglutide was favoured, with daily oral or weekly injectable administration being important attributes.⁴⁴ One study showed that patients preferred using either medication as second line treatment and preferred an oral, once daily treatment rather than injectable alternatives.⁴² Certainty of evidence was very low, downgraded because of very serious risk of bias and serious inconsistency, indirectness, and imprecision.

GLP-1 RA versus thiazolidinediones

Four observational studies compared GLP-1 RA and thiazolidinediones and provided very low certainty evidence of a consistent preference for GLP-1 RA.^44,46,50,51 Certainty was downgraded for very serious risk of bias, indirectness, and imprecision.

Other treatments compared with GLP-1 RA

Three observational studies compared GLP-1 RA with other glucose lowering treatments.^22,50,51 One study found that liraglutide was preferred over rosiglitazone and glimepiride, primarily due to weight loss, although when given by oral administration, rosiglitazone and glimepiride were preferred.²² Two additional studies comparing multiple drug classes showed that preferences were driven by medication attributes rather than by drug class, with influential factors being greater glycaemic efficacy, absence of hypoglycaemia, cardiovascular benefit, weight effects, and oral administration; no single class was consistently preferred.^50,51 Certainty of evidence for each medication class compared with GLP-1 RA was very low.

Comparison between medications for GLP-1 RA

Eight observational studies compared dulaglutide and semaglutide using different formulations and delivery routes.^{34–37,39,41,44,49} Preferences depended on the attributes presented. When oral once daily semaglutide was compared with once weekly injectable dulaglutide, semaglutide was preferred in three studies.^37,39,41 When once weekly injectable semaglutide was compared with once weekly injectable dulaglutide, dulaglutide was preferred in four of five studies, largely because of easier device use.^{34,39,41,44,49} One study incorporating clinical attributes for both medications favoured semaglutide, primarily caused by a greater reduction in HbA_1c, weight, and cardiovascular risk effects, and despite a higher rate of nausea.³⁷ Certainty of evidence was very low, downgraded for very serious risk of bias and indirectness, and for serious imprecision.

One observational study presented treatment processes and clinical data associated with once weekly injectable liraglutide and once daily oral semaglutide and provided very low certainty evidence for a preference for semaglutide.³⁶

Four observational studies compared liraglutide and exenatide and provided low to very low certainty evidence favouring once daily liraglutide over twice daily exenatide, but once weekly exenatide over once daily liraglutide.^21,22,29,30 Attributes generally determining preference were better efficacy, less frequent side-effects, and less frequent dosing.

Three observational studies comparing once daily liraglutide and once weekly injectable dulaglutide provided low to very low certainty evidence of a consistent preference for dulaglutide, primarily driven by dosing frequency and delivery system.^24,27,32

Five observational studies evaluated preferences for GLP-1 RA injection devices and found that preference declines as administration requirements (eg, needle handling, reconstitution, and waiting) increase.^{29,32,36,44,49} Once weekly devices and those requiring fewer preparation steps and where dose selection and delivery was clear were consistently preferred. Certainty of evidence was very low, downgraded for very serious indirectness, and serious risk of bias and imprecision.

SGLT-2i versus DPP-4i

Seven studies (randomised interventional and observational) compared SGLT-2i with DPP-4i, and showed a general preference for SGLT-2i, driven by perceived cardiovascular, kidney, and survival benefits, weight reduction, and glycaemic efficacy.^{38,43,44,46–48,51} Cost and risk of urogenital infections were also important considerations. Certainty of evidence was very low, downgraded for serious indirectness and imprecision.

SGLT-2i versus thiazolidinediones

Four observational studies comparing SGLT-2i to thiazolidinediones provided very low certainty evidence consistently favouring SGLT-2i.^44,46,50,53

SGLT-2i versus insulin

Two observational studies comparing SGLT-2i with insulin provided very low certainty evidence favouring SGLT-2i, primarily driven by its cardiovascular benefits.^46,51 Certainty of evidence was downgraded for very serious indirectness and imprecision, and serious risk of bias.

Discussion

Principal findings

In this updated systematic review, we synthesised the evidence on the values, preferences, and perceived treatment burden of adults with type 2 diabetes when initiating GLP-1 RA, SGLT-2i, tirzepatide, and finerenone. Across 34 studies, no high certainty evidence showed a consistent patient preference for one drug class over another. Instead, preferences were driven primarily by medication attributes; particularly, glycaemic efficacy, cardiovascular and kidney benefit, and simplicity of administration, rather than by drug class itself.

SGLT-2i and GLP-1 RA were generally preferred over insulin, thiazolidinediones, and DPP-4i, although, findings were variable for GLP-1 RA and DPP-4i. Evidence did not establish a clear preference between SGLT-2i and GLP-1 RA. When choosing between injection devices (pertinent to various GLP-1 RA and to insulin), patients favoured ease of use, reduced frequency of administration, and reliable dose delivery. Limited evidence supported tirzepatide over semaglutide, primarily due to HbA_1c reduction and lower hypoglycaemia risk. No studies assessed preferences of finerenone.

Comparison with previous evidence

These findings are consistent with our prior 2021 systematic review, which identified limited and very low certainty evidence and highlighted the importance of treatment burden and delivery characteristics in shaping patient preferences. The present update expands the evidence and incorporates newer treatments and patient important outcomes, particularly cardiovascular benefits, reflecting the evolving paradigm of diabetes care. While more studies increasingly included cardiovascular outcomes in drug profiles, kidney outcomes remained rarely presented, despite their relevance to patients and clinicians.

Overall, our findings reinforce that patient preferences are heterogeneous and context dependent, and that simplifying treatment delivery may be as influential as improving clinical efficacy.

Implications for clinical practice and guidelines

The absence of high certainty evidence supporting a uniform preference between SGLT-2i and GLP-1 RA underscores the importance of shared decision making. Clinicians should avoid assuming that one class is universally preferred and should instead engage patients in discussions that consider treatment goals, anticipated benefits and harms, route and frequency of administration, treatment burden, and individual circumstances.

For guideline developers, these findings highlight the need to explicitly acknowledge variability in patient preferences and to avoid strong recommendations based solely on assumed patient values, particularly where the balance of benefits and harms and burdens are close. Decision aids that transparently present benefits, harms, and treatment burden may also help to align treatment choice with patient priorities.

Strengths and limitations

Our review has several strengths. We used a sensitive and previously validated search strategy to identify eligible trials. Furthermore, we followed rigorous methodological standards in conducting the review where study selection, data extraction, and risk of bias evaluation were conducted independently and in duplicate. Risk of bias was evaluated using a dedicated tool for values and preferences studies endorsed by GRADE, and the GRADE approach was used to evaluate overall certainty of evidence for each comparison.

However, we acknowledge important limitations. Most included studies were observational and at high risk of bias, and certainty of evidence for all comparisons was low or very low. Drug profiles were often incomplete or not based on the best available evidence, many studies were industry funded, and outcomes of greatest importance to patients remain only partially represented even in the most recently conducted studies. Attributes informing patient preferences may, therefore, be affected by selective framing. As a result, our findings should be interpreted cautiously. Furthermore, our review focused on studies that explicitly elicited preferences regarding diabetes treatments and their attributes. Consequently, qualitative studies exploring patients' experiences and perceptions through interviews and focus groups were generally outside of the scope of our eligibility criteria.

Implications for future research

Higher quality studies are needed to evaluate patient values and preferences using standardised, evidence based drug profiles that incorporate outcomes that patients care about. These outcomes include cardiovascular and kidney outcomes. Future studies should prioritise comparisons between guideline recommended treatments, particularly SGLT-2i, once weekly GLP-1 RA, and emerging treatments such as tirzepatide and finerenone. Furthermore, they should explicitly examine how preferences vary by patient characteristics and treatment experience. Future research should prioritise the use of validated elicitation tools, more representative patient samples, and transparently reported attributes and scenarios presented to participants. Greater standardisation in study design and outcome reporting, and efforts to minimise methodological bias would improve comparability across studies and strengthen confidence in the evidence. Addressing these limitations and increasing use of independent funding sources would enhance credibility of future research and improve its applicability to real-world decision making. Future research might also benefit from integrating quantitative preference elicitation methods with qualitative approaches, to better understand motivations, experiences, and contextual factors underlying patients' treatment preferences.

Conclusions

Current evidence does not support a clear patient preference for any single glucose lowering treatment. Instead, low or very low certainty evidence suggests that adults with type 2 diabetes prioritise effective glycaemic control, cardiovascular benefit, and simple, reliable treatment delivery. These findings reinforce the need for individualised treatment decisions that are grounded in shared decision making and informed by the best available evidence on both clinical outcomes and patient preferences. Future research should incorporate high quality mixed-methods approaches to allow for a comprehensive foundation for guideline development and shared decision making.

Data availability statement

Data are available upon reasonable request. All data relevant to the study are included in the article or uploaded as supplementary information.