How many patients were actually randomized to metformin in UKPDS 34?

753 patients received metformin as primary therapy. An additional 537 were in the "metformin added to sulfonylurea" sub-study. The total trial population of 1,704 includes the conventional therapy comparator arm (n = 411) and other subgroups, making the metformin-specific evidence base smaller than the headline number suggests.

Was UKPDS 34 blinded?

No. The trial used an open-label design where both patients and clinicians knew treatment assignment. Endpoint adjudication was performed by a committee, but the lack of double-blinding is a recognized limitation, particularly for procedure-based endpoints in the composite.

Did the trial adjust for multiple comparisons?

The original publication did not apply formal multiplicity correction across its primary comparisons. The headline result (p = 0.002) would likely survive Bonferroni adjustment, but secondary findings, including the paradoxical harm signal with sulfonylurea combination, would not.

Why was metformin only tested in overweight patients?

The UKPDS investigators hypothesized that metformin's insulin-sensitizing mechanism would be most relevant in patients with obesity-driven insulin resistance. This was a reasonable clinical hypothesis but limited the direct evidence base to patients above 120% ideal body weight.

Has any subsequent RCT replicated the mortality benefit?

No large-scale RCT has replicated the cardiovascular mortality reduction seen in UKPDS 34 for metformin monotherapy in a newly diagnosed population. The post-trial monitoring study showed persistent benefits, and observational data is broadly supportive, but a modern confirmatory CVOT has never been conducted.

What happened with the metformin-plus-sulfonylurea mortality signal?

UKPDS 34 reported a 96% increase in diabetes-related death when metformin was added to sulfonylurea therapy. Subsequent meta-analyses did not replicate this finding. The current clinical consensus treats it as a likely chance finding in a small subgroup, and metformin-sulfonylurea combinations remain widely prescribed.

How did background therapy changes affect the results?

The trial spanned 1977 to the late 1990s. Statins, ACE inhibitors, and aspirin became standard during this period. The conventional arm's outcomes partly reflect an era before these therapies were routine, which may have exaggerated the relative benefit of metformin.

Is UKPDS 34 still relevant given newer diabetes drugs?

Yes. UKPDS 34 remains the only RCT showing mortality benefit for metformin monotherapy in newly diagnosed type 2 diabetes. SGLT2 inhibitors and GLP-1 receptor agonists have demonstrated cardiovascular benefits in higher-risk populations, but metformin's cost, safety record, and this trial's long-term data keep it as a guideline-recommended first-line agent.

Did UKPDS 34 include diverse racial or ethnic populations?

The trial was approximately 81% white British. Asian Indian and Afro-Caribbean patients comprised roughly 18% combined. This limited diversity means the cardiovascular mortality benefit has not been confirmed by RCT in populations where type 2 diabetes prevalence is highest globally.

What are the lactic acidosis concerns with metformin?

UKPDS 34 reported no cases of lactic acidosis, but the trial excluded patients with renal impairment, hepatic disease, and heart failure. The FDA relaxed metformin's renal contraindication in 2016, now permitting use down to eGFR 30 mL/min/1.73m², based on accumulated safety data showing the risk is far lower than initially feared from the phenformin era.

Honest Criticisms and Limitations of the UKPDS 34 Trial

By HealthRX Medical Team

Published May 25, 2026Updated May 25, 2026Last reviewed May 25, 2026

At a glance

| Parameter | Detail | |---|---| | N | 1,704 overweight patients (of 4 to 075 in the broader overweight UKPDS cohort) | | Intervention | Metformin (up to 2 to 550 mg/day) | | Comparator | Conventional (diet-only) therapy; secondary comparisons with sulfonylurea/insulin | | Duration | Median 10.7 years follow-up | | Primary endpoint | Any diabetes-related clinical endpoint (sudden death, hyperglycemia/hypoglycemia requiring hospital admission, MI, angina, heart failure, stroke, renal failure, amputation, vitreous hemorrhage, retinal photocoagulation, blindness, cataract extraction) | | Key result | 32% risk reduction in diabetes-related endpoints vs. conventional therapy (p = 0.002) |

The Trial That Built Metformin's Throne

The UKPDS 34 publication appeared in the Lancet in 1998 and immediately reshaped diabetes care worldwide. Metformin reduced diabetes-related endpoints by 32%, all-cause mortality by 36%, and myocardial infarction by 39% in overweight patients with newly diagnosed type 2 diabetes. No sulfonylurea or insulin regimen in the parallel UKPDS 33 trial achieved comparable mortality benefits despite similar glycemic control. The implication was clear: metformin did something beyond glucose lowering.

But the trial was designed in 1977, enrolled its first patient in 1977, and published two decades later. Its methods reflect an era before CONSORT reporting standards, before mandatory trial registration, and before the cardiovascular outcome trial (CVOT) framework the FDA now expects for glucose-lowering drugs. The criticisms below are not contrarian posturing. They are the specific methodological realities that every clinician prescribing metformin should understand.

Enrollment Bias and Population Restrictions

UKPDS 34 restricted randomization to metformin only among patients whose ideal body weight exceeded 120%. Of the 4,075 overweight patients in the UKPDS overweight stratum, just 753 were randomized to metformin. The remaining patients were assigned to conventional therapy (n = 411) or to sulfonylurea/insulin arms in the main UKPDS 33 protocol.

This creates two problems:

The metformin arm was small. With 753 patients, even a median 10.7-year follow-up yields modest event counts. The primary endpoint aggregate (diabetes-related endpoints) accumulated 128 events in the metformin group vs. 207 events expected under the conventional arm's rate. Individual components, particularly stroke (fatal and non-fatal), did not reach significance on their own.
Generalizability to non-overweight patients was assumed, not demonstrated. The trial explicitly excluded normal-weight type 2 diabetes from metformin randomization. Subsequent guideline bodies, including the ADA Standards of Care, extended metformin to all BMI categories based on mechanism-of-action reasoning and observational data, not on RCT evidence from UKPDS 34 itself.

The HealthRX Limitation Taxonomy for UKPDS 34

The following framework organizes the trial's limitations into five domains. Each domain scores the concern's impact on how much confidence a clinician should place in the headline result.

| Domain | Specific Concern | Severity | |---|---|---| | Design | Open-label assignment (no blinding of patients or clinicians) | Moderate | | Design | No placebo arm; "conventional" = diet only, which drifted over 20 years | High | | Population | Restricted to overweight, newly diagnosed T2D; results extrapolated beyond this | High | | Population | 81% white British; minimal ethnic diversity | Moderate | | Statistical | Multiple primary comparisons without pre-specified alpha adjustment | High | | Statistical | Composite endpoint mixes hard endpoints (MI, death) with soft (photocoagulation) | Moderate | | Follow-up | 20-year study span means background care changed substantially (statins, ACE inhibitors became standard) | Moderate | | Conflict/Context | Funded partly by pharmaceutical sponsors including Bristol-Myers Squibb (metformin manufacturer at the time) | Low-Moderate |

Open-Label Design and Ascertainment Bias

UKPDS 34 was open-label. Patients and clinicians knew the treatment assignment. For a trial with "soft" endpoints in its composite (retinal photocoagulation, cataract extraction), knowledge of assignment could influence the threshold at which a clinician refers for a procedure. Hard endpoints like myocardial infarction and death are less susceptible to this bias, and metformin's mortality reduction (36%, p = 0.011) is reassuring precisely because mortality is difficult to misclassify.

Still, the original publication does not describe blinded endpoint adjudication for the composite components. The UKPDS group later clarified that an endpoint committee reviewed events, but the absence of patient or clinician blinding remains a limitation by modern CVOT standards, where double-blinding and independent adjudication committees are expected.

The Multiplicity Problem

This is the criticism that generated the most post-publication debate. UKPDS 34 made three primary comparisons for metformin:

Metformin vs. conventional therapy (the headline comparison)
Metformin vs. sulfonylurea/insulin (the "added to sulfonylurea" sub-study)
Metformin vs. chlorpropamide, glibenclamide, or insulin individually

The trial reported p-values without formal correction for multiple testing. The 32% reduction carried p = 0.002, which would survive a Bonferroni correction. But the secondary finding that adding metformin to sulfonylurea increased mortality (96% increase in diabetes-related death, p = 0.039) would not survive even modest adjustment. This paradoxical harm signal created confusion that persisted for years, with subsequent analyses and the UKPDS post-trial monitoring study eventually arguing the combination signal was likely a chance finding in a small subgroup (n = 537).

The core tension: the trial's statistical framework was designed in the late 1970s. Adaptive designs, hierarchical testing procedures, and gate-keeping strategies that modern trials employ were not standard practice.

The Composite Endpoint: Hard and Soft Mixed Together

The primary endpoint grouped 13 different clinical events into a single composite. This is standard practice in cardiovascular trials, but UKPDS 34's composite is unusually heterogeneous. It includes:

Events patients fear most: MI, stroke, sudden death, amputation
Events with real morbidity but lower mortality weight: renal failure requiring dialysis, heart failure
Procedure-based endpoints: retinal photocoagulation, cataract extraction

Procedure-based endpoints are influenced by access to care, local practice patterns, and the era in which they occurred. A patient in 1985 received different ophthalmologic screening than one in 1995. Over a 20-year enrollment window, these shifts introduce noise that pure clinical events (death, MI) do not.

When UKPDS 34's results are restricted to the three pre-specified hard endpoints (diabetes-related death, all-cause mortality, MI), metformin still shows benefit, but the confidence intervals widen and the p-values for individual components become borderline. The composite was necessary for statistical power. That necessity is itself a limitation.

Ethnic and Geographic Homogeneity

UKPDS enrolled patients across 23 UK hospital-based clinics between 1977 and 1991. The trial population was approximately 81% white European, 10% Asian Indian, and 8% Afro-Caribbean. Type 2 diabetes prevalence, phenotype, and cardiovascular risk profiles differ substantially across ethnic groups. South Asian patients, for example, tend to develop insulin resistance at lower BMI thresholds.

Modern metformin prescribing spans every ethnicity worldwide. The evidence base for that breadth rests on UKPDS 34 (predominantly white British), observational cohorts, and mechanism-of-action extrapolation. No subsequent RCT of comparable size has replicated the cardiovascular mortality benefit of metformin in a more diverse population.

Background Therapy Drift Over Two Decades

Patients entered UKPDS between 1977 and 1991. During the trial's follow-up period, statins became available (lovastatin approved 1987, simvastatin widely prescribed by the mid-1990s), ACE inhibitors became standard for diabetic nephropathy, and aspirin for secondary prevention became routine. The conventional (diet-only) arm's outcomes reflect an era when these therapies were either unavailable or inconsistently used.

This raises a question that cannot be resolved from the trial data: would metformin still show a 32% relative risk reduction against a "conventional" arm that included statins and antihypertensives from the start? The UKPDS post-trial monitoring showed persistent mortality benefits 10 years after trial end, even as background therapy improved. This is encouraging but not the same as a de novo trial in the statin era.

What Post-Publication Commentary Raised

Several letters to the Lancet and subsequent editorials highlighted concerns that the original publication did not fully address:

Lactic acidosis signal. Phenformin (a related biguanide) was withdrawn for fatal lactic acidosis. UKPDS 34 reported no cases, but the trial excluded patients with renal impairment (serum creatinine >175 µmol/L), hepatic disease, and heart failure. Real-world prescribing frequently includes these populations. The FDA metformin label was updated in 2016 to relax the renal contraindication, partly because UKPDS and observational data showed minimal lactic acidosis risk.
The sulfonylurea combination paradox. As noted above, the finding that metformin added to sulfonylurea increased diabetes-related death generated substantial concern. A meta-analysis by Rao et al. (2008) examined this signal across multiple datasets and concluded the UKPDS subgroup finding was not replicated, though residual uncertainty remains.
Intent-to-treat dilution. Over 10.7 years of median follow-up, substantial crossover occurred. Patients in the conventional arm eventually received pharmacotherapy as glycemic targets were not met. This dilutes the between-group difference and may underestimate metformin's true effect, or may simply reflect the reality of treating a progressive disease.

What This Means for Prescribing Today

None of these limitations suggest metformin is ineffective. The aggregate evidence from UKPDS 34, the post-trial monitoring study, decades of observational data, and the drug's favorable safety profile at generic cost supports its position as first-line therapy per ADA and EASD guidelines.

What the limitations do suggest: the magnitude of cardiovascular benefit (32-39% reductions) may be inflated by open-label design, composite endpoint heterogeneity, and a comparator arm that lacked modern background therapy. Clinicians should hold the direction of benefit with confidence and the exact effect size with appropriate humility.

Frequently asked questions

›

References

UK Prospective Diabetes Study (UKPDS) Group. Effect of intensive blood-glucose control with metformin on complications in overweight patients with type 2 diabetes (UKPDS 34). Lancet. 1998;352(9131):854-865. PubMed
Holman RR, Paul SK, Bethel MA, Matthews DR, Neil HA. 10-year follow-up of intensive glucose control in type 2 diabetes. N Engl J Med. 2008;359(15):1577-1589. PubMed
Rao AD, Kuhadiya N, Reynolds K, Fonseca VA. Is the combination of sulfonylureas and metformin associated with an increased risk of cardiovascular disease or all-cause mortality? Diabetes Care. 2008;31(8):1672-1678. PubMed
American Diabetes Association Professional Practice Committee. Standards of Care in Diabetes, 2023. Diabetes Care. 2023;46(Suppl 1). PubMed
FDA. Metformin hydrochloride prescribing information (revised 2017). FDA Label
Davies MJ, D'Alessio DA, Fradkin J, et al. Management of hyperglycemia in type 2 diabetes, 2018: a consensus report by the ADA and EASD. Diabetes Care. 2018;41(12):2669-2701. PubMed