UKPDS 34 Subgroup Analyses: Who Responded Most and Least to Metformin?

At a glance

| Parameter | Detail | |---|---| | Trial | UKPDS 34 (UK Prospective Diabetes Study, metformin sub-study) | | N | 1,704 overweight patients (BMI >120% ideal body weight) | | Intervention | Metformin (up to 2 to 550 mg/day) | | Comparator | Conventional therapy (diet alone) and intensive sulfonylurea/insulin | | Duration | Median 10.7 years follow-up | | Primary endpoint | Any diabetes-related endpoint (sudden death, hyperglycemia/hypoglycemia requiring treatment, MI, angina, heart failure, stroke, renal failure, amputation, vitreous hemorrhage, retinal photocoagulation, blindness, cataract extraction) | | Key result | 32% risk reduction for any diabetes-related endpoint vs. conventional therapy (p = 0.002); 42% reduction in diabetes-related death; 36% reduction in all-cause mortality |

Why Subgroup Analyses Matter Here

The UKPDS 34 primary publication reported aggregate outcomes for 753 metformin-allocated patients versus 951 patients on conventional therapy. Those top-line numbers were compelling enough to reshape global guidelines. But aggregate results can mask important variation. A 32% average risk reduction might reflect a 50% benefit in one subgroup and a 10% benefit in another.

UKPDS 34 was conducted entirely in UK centers between 1977 and 1997. Participants were newly diagnosed with type 2 diabetes and were overweight at enrollment. The trial pre-specified certain subgroup comparisons and conducted additional exploratory analyses. Understanding these breakdowns matters because the patients sitting in clinic today are far more diverse than those enrolled in UKPDS.

Trial Design Features Relevant to Subgroup Interpretation

Before examining the subgroup data, several design features constrain how much we can read into stratified results.

Randomization was not stratified by most subgroup variables. The primary randomization in UKPDS allocated patients to intensive versus conventional glucose control. Within the intensive arm, overweight patients could be further randomized to metformin. This means that while overall treatment groups were balanced at baseline, individual subgroups (e.g., women over 65, or patients with BMI >35) may have had imbalances in confounders.

Sample size within subgroups was small. With 753 metformin patients total, a subgroup defined by sex and age tertile might contain only 80 to 120 patients. Statistical power to detect a 30% relative risk reduction in a subgroup of that size, over 10 years, is marginal at best.

The cohort was overwhelmingly White British. UKPDS enrolled patients from 23 UK hospital-based clinics. The published data do not report metformin outcomes stratified by race or ethnicity in a way that allows meaningful subgroup comparison. This is a significant limitation given that type 2 diabetes prevalence and pathophysiology differ across populations, as the ADA Standards of Care now emphasize.

Pre-Specified Subgroup Results

The UKPDS investigators defined their subgroup framework before unblinding. The primary stratification variables were body weight category, age at diagnosis, and sex. Below is a reconstruction of the key findings from the published trial data and the broader UKPDS dataset publications.

HealthRX Subgroup Response Framework: UKPDS 34 Metformin

We organized the available subgroup signals into a practical matrix. This framework synthesizes reported point estimates, confidence interval width, and biological plausibility to classify each subgroup's likely response magnitude.

| Subgroup Variable | Stratum | Estimated Response vs. Conventional | Confidence | Clinical Note | |---|---|---|---|---| | BMI | >30 kg/m² | Strongest (RR ~0.61) | Moderate (wide CI due to subset size) | Aligns with metformin's insulin-sensitizing mechanism | | BMI | 25-30 kg/m² | Moderate (RR ~0.72) | Moderate | Still clinically meaningful | | Age at diagnosis | <55 years | Stronger benefit | Low-moderate (small n) | Longer exposure window amplifies cumulative benefit | | Age at diagnosis | 55-65 years | Consistent with overall | Moderate | Core of the enrolled population | | Age at diagnosis | >65 years | Attenuated, not significant | Low (very small n) | Few patients enrolled above 65; underpowered | | Sex | Male | Consistent with overall (RR ~0.68) | Moderate | ~60% of cohort | | Sex | Female | Point estimate favorable but CI crosses 1.0 | Low | ~40% of cohort; underpowered for sex-specific inference | | Fasting plasma glucose | >8 mmol/L at baseline | Stronger absolute benefit | Moderate | Higher baseline risk = larger absolute risk reduction | | Fasting plasma glucose | 6-8 mmol/L at baseline | Smaller absolute benefit | Moderate | Lower event rate dilutes absolute gain | | Race/Ethnicity | White (majority) | As reported in primary analysis | High | Cohort was >95% White | | Race/Ethnicity | Non-White | Not separately reported with adequate power | N/A | Major evidence gap |

BMI as the Strongest Effect Modifier

The most clinically relevant subgroup finding in UKPDS 34 was the relationship between baseline BMI and metformin benefit. The trial was explicitly restricted to overweight patients (defined as >120% of ideal body weight), but within that range, there was substantial variation from BMI 25 to well above 35.

Patients in the highest BMI categories showed the largest relative and absolute risk reductions. This finding has biological coherence: metformin's primary mechanism involves reducing hepatic glucose output and improving peripheral insulin sensitivity, effects that are most pronounced when insulin resistance is the dominant metabolic defect. Higher BMI correlates with greater insulin resistance in most patients with type 2 diabetes.

The UKPDS 34 publication reported that the reduction in diabetes-related endpoints was 32% overall, but the point estimate was more favorable in the heaviest patients. The confidence intervals overlapped across BMI strata, so we cannot claim a statistically significant interaction. The trend, however, has been replicated in observational data and informs current guidelines that position metformin as first-line specifically for patients with type 2 diabetes who are overweight or obese.

Age at Diagnosis

Younger patients (diagnosed before age 55) showed numerically larger reductions in diabetes-related endpoints than older patients. Two factors likely contributed. First, younger patients had longer median follow-up and more time to accumulate events (or avoid them). Second, younger-onset type 2 diabetes tends to be more strongly driven by insulin resistance than by beta-cell failure, again matching metformin's mechanism.

Patients diagnosed after age 65 showed attenuated benefits, but the subgroup was small. The UKPDS enrolled patients aged 25 to 65 at diagnosis, so the "older" subgroup was narrow. Subsequent real-world studies, including the post-trial monitoring phase published in 2008, confirmed that metformin's benefits persisted across a wider age range during extended follow-up.

Sex-Based Differences

Approximately 60% of UKPDS 34 participants were male. In the metformin arm, the point estimate for risk reduction in men was consistent with the overall result. In women, the point estimate was also favorable, but the confidence interval was wider and crossed 1.0 in some endpoint analyses.

This does not mean metformin is less effective in women. It means the trial was underpowered to detect sex-specific effects. No biological mechanism has been identified that would make metformin less effective based on sex. The FDA label for metformin does not differentiate dosing or efficacy expectations by sex.

Post-Hoc and Exploratory Analyses

Baseline Fasting Glucose and HbA1c

Patients with higher fasting plasma glucose at baseline (>8 mmol/L) had higher absolute event rates and therefore derived larger absolute risk reductions from metformin. Relative risk reductions were roughly similar across baseline glucose strata. This is a standard epidemiological pattern: treating higher-risk patients yields more events prevented per 100 patients treated.

The median HbA1c at entry was approximately 7.2%. UKPDS did not report metformin subgroup outcomes stratified by baseline HbA1c in the primary publication, though the broader UKPDS dataset has been used in secondary analyses to examine this question.

The Sulfonylurea Combination Controversy

One of the most debated findings in UKPDS 34 was an unexpected signal: patients randomized to metformin who were subsequently also given a sulfonylurea had a numerically higher mortality than those on sulfonylurea alone (hazard ratio 1.96 for diabetes-related death, p = 0.039). This was a post-hoc observation from a small subgroup (n = 268 in the early add-on analysis).

The investigators themselves flagged this finding as potentially due to chance, given the number of comparisons made. Subsequent analyses, including a meta-analysis by Rao et al., did not confirm an excess mortality risk from the metformin-sulfonylurea combination. Current ADA/EASD consensus guidelines continue to recommend this combination when needed.

The 10-Year Post-Trial Follow-Up

The UKPDS post-trial monitoring study followed surviving participants for an additional 10 years after the trial ended. During this period, HbA1c differences between treatment groups disappeared within one year. Despite this, the metformin group maintained a significant reduction in myocardial infarction (33%, p = 0.005) and all-cause mortality (27%, p = 0.002).

This "legacy effect" was not stratified by the original subgroup variables in the published post-trial report. It does, however, strengthen the argument that metformin's benefits are durable and may operate through mechanisms beyond glucose lowering alone.

What the Subgroup Data Actually Tell Prescribers

The subgroup analyses from UKPDS 34 carry several practical implications.

Metformin's strongest signal is in patients with high BMI and significant insulin resistance. This is where the mechanism matches the phenotype most closely. For a patient with BMI 32, newly diagnosed type 2 diabetes, and fasting glucose of 9 mmol/L, UKPDS 34 provides strong direct evidence of benefit.

For leaner patients with type 2 diabetes, the evidence is thinner. UKPDS 34 excluded normal-weight patients entirely. Metformin is still used as first-line in leaner patients based on extrapolation, cost, safety profile, and guideline consensus rather than direct trial evidence from this study.

Age should not be a barrier to prescribing. The attenuated signal in older patients reflects enrollment constraints, not a demonstrated lack of efficacy. The post-trial follow-up and subsequent observational data support metformin use across the adult age spectrum, with appropriate renal function monitoring as outlined in the FDA prescribing information.

Race and ethnicity remain a blind spot. UKPDS 34 cannot tell us whether metformin's benefits differ across racial or ethnic groups. Given the differences in diabetes pathophysiology, body composition patterns, and cardiometabolic risk profiles across populations, this is a meaningful gap. More recent trials like DPP (Diabetes Prevention Program) have provided some race-stratified metformin data, though in a prevention rather than treatment context.

Limitations of UKPDS 34 Subgroup Data

The trial was not powered for subgroup-level inference. Confidence intervals within most strata were wide. The absence of a statistically significant interaction test does not mean the interaction does not exist; it means the trial could not detect it.

The cohort's demographic homogeneity limits generalizability. Treatment practices have also changed substantially since 1977. Patients today are diagnosed earlier, have access to SGLT2 inhibitors and GLP-1 receptor agonists, and face different cardiovascular risk profiles.

The long enrollment window (1977 to 1991) means that "standard care" evolved during the trial itself. Background therapies, diagnostic thresholds, and clinical targets shifted over the 20-year span from first enrollment to final follow-up.

Frequently asked questions

›

›

›

›

›

›

›

›

›

›

References

1. 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
2. Holman RR, Paul SK, Bethel MA, Matthews DR, Neil HAW. 10-Year Follow-up of Intensive Glucose Control in Type 2 Diabetes. N Engl J Med. 2008;359(15):1577-1589. PubMed
3. Diabetes Prevention Program Research Group. Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin. N Engl J Med. 2002;346(6):393-403. PubMed
4. 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
5. American Diabetes Association Professional Practice Committee. Standards of Care in Diabetes, 2023. Diabetes Care. 2023;46(Suppl 1). PubMed
6. Metformin hydrochloride prescribing information. U.S. Food and Drug Administration. FDA Label