What 869 trials say about diabetes drugs

A living BMJ review maps the benefits and harms of 63 drugs, showing why heart, kidney and weight priorities matter more than a single league table.

✓ Verified Source BMJ living systematic review and network meta-analysis, cross-checked with PubMed/Europe PMC and Crossref ⚑ Diabetes drugs

The 60-second version

A living BMJ review of 869 randomized trials shows that type 2 diabetes medicines have different strengths and harms, so treatment choice should follow a person's risk profile and priorities.

Key points

  • SGLT-2 inhibitors and GLP-1 receptor agonists have established cardiovascular and kidney benefits, while finerenone benefits selected patients with chronic kidney disease.
  • Tirzepatide produced the largest average weight reduction in the network, but weight is only one outcome and gastrointestinal harms matter.
  • Network meta-analysis mixes direct and indirect comparisons, and absolute benefits depend strongly on baseline risk.
  • Drug-specific harms include genital infection, ketoacidosis, gastrointestinal events, hyperkalaemia, fracture, heart-failure hospitalization and hypoglycaemia.

Verdict. Use the evidence to structure a clinician conversation, not to self-prescribe or chase a universal winner.

Bottom lineThe review is a map, not a single ranking

A living BMJ review combined 869 randomized trials involving 493,168 adults with type 2 diabetes. It found no one drug that wins every outcome: cardiovascular risk, kidney disease, weight, hypoglycaemia and adverse effects point toward different choices.

869randomized trials
493,168participants
63drugs across 13 classes
26benefit and harm outcomes

MethodHow the evidence network works

The review included randomized trials lasting at least 24 weeks. A network meta-analysis combines direct head-to-head trials with indirect comparisons through a shared comparator. That allows a broad map, but it also means certainty differs from one comparison to another.

The authors used GRADE to rate confidence and a random-effects model to allow for differences across studies. The search for this version ran through 31 July 2024; the project plans updates at least twice a year.

BenefitsHeart, kidney and weight outcomes do not align perfectly

Moderate- to high-certainty evidence confirmed cardiovascular and kidney benefits for SGLT-2 inhibitors and GLP-1 receptor agonists. Finerenone showed benefit for patients who already have chronic kidney disease. The absolute gain varies substantially with a person's baseline cardiovascular and kidney risk.

Cardiovascular and kidney riskSGLT-2 inhibitors and GLP-1 receptor agonists have established benefits; absolute gains are larger when baseline risk is higher.
Chronic kidney diseaseFinerenone adds evidence for selected patients with established CKD, alongside its potassium risk.
WeightTirzepatide had the largest average reduction in the network; orforglipron followed, but its evidence was less certain.
Severe hypoglycaemiaSulfonylureas, insulin and DPP-4 inhibitors probably increased risk.
−8.63 kgtirzepatide mean weight difference; moderate certainty
−7.87 kgorforglipron mean difference; low certainty

Those weight estimates are averages across a trial network, not forecasts for one person. They also do not settle other priorities such as glucose control, cardiovascular disease, kidney function, tolerability, cost or access.

HarmsEach class carries a different trade-off

SGLT-2 inhibitorsMore genital infections (OR 3.29) and diabetic ketoacidosis (OR 2.08); amputations probably increased modestly (OR 1.27).
Tirzepatide and GLP-1 drugsProbably more severe gastrointestinal events; the largest estimate was for tirzepatide (OR 4.21).
FinerenoneMore severe hyperkalaemia (OR 5.92), making potassium and kidney monitoring relevant.
ThiazolidinedionesMore major osteoporotic fractures and probably more hospitalization for heart failure.

LimitsWhat the review cannot settle

  • Not every drug pair was compared directly; some estimates come from indirect links in the network.
  • Trials differed in populations, follow-up and outcome definitions, even though statistical models account for heterogeneity.
  • Evidence for neuropathy and visual impairment was low or very low certainty.
  • Whether GLP-1 drugs reduce dementia remained uncertain: OR 0.92, 95% CI 0.83 to 1.02, with low certainty.
  • The current search cutoff was July 2024, so a living review still requires later updates.
The best-supported choice depends on the outcome a patient most needs to improve and the harm they most need to avoid.

Next stepTurn the evidence into the right questions

For the next clinical review, bring a short list: cardiovascular disease or heart failure, kidney function, desired weight change, previous adverse effects, hypoglycaemia risk, other medicines, cost and access. The review makes those trade-offs easier to discuss; it does not replace individualized care.