Type 2 Diabetes: History of Treatment Over Decades

Type 2 diabetes treatment did not evolve as a straight line of better drugs replacing worse ones. It evolved as a shift in the question clinicians were trying to answer. For most of the twentieth century, the question was "how do we lower blood glucose." Since roughly the mid-2010s, the operative question has become "which comorbidity does this person carry, and which drug class has trial evidence for that specific risk." That reframing, more than any single molecule, is the part of this history a reader should take away and test against current prescribing.

This is educational and historical content. It is not a treatment recommendation, and no dosing or drug-selection decision should be made from it without a clinician who has reviewed the individual's kidney function, cardiovascular history, and current medication list.

Direct answer

Type 2 diabetes drug therapy has moved through five broad, overlapping eras: dietary restriction before the 1950s, sulfonylureas from 1955 onward, metformin's rise to first-line status culminating in its 1994 US approval, a wave of thiazolidinedione and incretin-based drugs from the late 1990s through the 2000s, and the current era of GLP-1 receptor agonists and SGLT2 inhibitors, which are the first antidiabetic classes with randomized-trial evidence for reducing cardiovascular death and slowing kidney disease progression rather than only lowering blood glucose. Current American Diabetes Association (ADA) guidance selects add-on therapy by comorbidity (cardiovascular disease, heart failure, chronic kidney disease) rather than by glucose level alone, which is a materially different clinical logic than the one used as recently as the early 2000s.

The pre-drug era: diet as the only tool (before 1950)

Before any pharmacological agent existed, clinicians managed what would now be classified as type 2 diabetes almost entirely through caloric and carbohydrate restriction. Very-low-calorie regimens popularized in the early twentieth century lowered blood glucose but caused significant malnutrition in some patients. The discovery of insulin in 1921 transformed care for type 1 diabetes almost immediately; its role in adult-onset diabetes was recognized more slowly, since many of these patients retained some beta-cell function and did not require insulin to survive.

The American Diabetes Association, founded in 1940, published early dietary guidance that emphasized weight reduction and carbohydrate limitation, with no pharmacological pathway beyond insulin for patients who could not reach control through diet alone. The organization's standards have been revised annually since; the current version is the reference point for present-day treatment decisions, available from the ADA's journal.

Sulfonylureas: the first oral drugs (1950s to 1970s)

The first oral antidiabetic class emerged from an observation in antibacterial sulfonamide research: one compound caused unexpected hypoglycemia. Carbutamide was introduced in Europe in 1955 as the first sulfonylurea, and tolbutamide reached the US market in 1957. Sulfonylureas stimulate pancreatic beta cells to secrete insulin independent of ambient glucose concentration, a mechanism that lowers blood sugar reliably but carries a real hypoglycemia risk that later drug classes were specifically designed to avoid.

The University Group Diabetes Program, a US trial reported around 1970, raised a signal that tolbutamide might be associated with excess cardiovascular mortality compared with insulin or placebo. The study's design was heavily debated in the years that followed, but the finding was consequential regardless of its methodological quality: it is the reason the FDA later required cardiovascular outcome data for new diabetes drugs, a policy that shaped every drug class discussed below. Readers researching this specific study should look for the primary publication and its later critiques rather than rely on secondary summaries, since accounts of its exact statistical findings vary.

Later sulfonylureas (glipizide and glyburide in 1984, glimepiride in 1995) offered better receptor selectivity and, in glimepiride's case, once-daily dosing with somewhat less hypoglycemia than glyburide. Sulfonylureas remain in use today, mainly on cost grounds; current ADA guidance ranks them below drug classes with proven cardiovascular or kidney benefit for patients who have those conditions.

Metformin: discovered in the 1950s, first-line since the late 1990s

Metformin descends from galegine, a compound found in the plant Galega officinalis, and was first described as a diabetes treatment by Jean Sterne in the 1950s. Its primary mechanism, activation of AMP-activated protein kinase in the liver, suppresses hepatic glucose production and improves peripheral insulin sensitivity without stimulating insulin secretion, which is why it does not cause hypoglycemia on its own.

Phenformin, a related biguanide, was withdrawn from the US market in 1977 after being linked to fatal lactic acidosis at rates far higher than metformin has ever produced. Metformin carries a black-box warning about lactic acidosis, but the risk in practice is concentrated almost entirely in patients with severely reduced kidney function; a Cochrane systematic review on this question exists and is the appropriate reference for exact incidence figures, which this article does not reproduce because the specific citation could not be verified against the source material provided.

The UK Prospective Diabetes Study (UKPDS), a large randomized trial published in the late 1990s, is widely credited as the pivotal evidence base for metformin in overweight patients with type 2 diabetes, showing reductions in diabetes-related complications and mortality compared with diet-only management over roughly a decade of follow-up. The exact percentage reductions commonly cited for UKPDS are well established in the diabetes literature, but this draft does not attach a specific numeric figure to an unverified link; an editor with primary-literature access should confirm the precise effect sizes before publication.

Metformin received FDA approval in the United States in 1994. Its kidney-function eligibility threshold was relaxed in a 2016 FDA label update, which reduced restrictions on metformin use in patients with reduced kidney function (confirm current label language before citing it as active guidance, since labels are revised over time). It remains the preferred initial agent in current ADA guidance for most patients without established cardiovascular or kidney disease, primarily on the basis of long safety experience, weight-neutral to modestly weight-favorable profile, and low cost.

Thiazolidinediones: a class that split into a caution and a survivor (1990s-2000s)

Thiazolidinediones (TZDs) activate PPAR-gamma receptors, improving insulin sensitivity in fat, muscle, and liver. Troglitazone reached the US market in 1997 and was withdrawn in 2000 after being linked to fulminant hepatic failure, a hepatotoxicity signal not shared by the two TZDs that followed it.

Rosiglitazone became one of the more consequential drug-safety episodes in diabetes care. A widely cited 2007 meta-analysis, published in the New England Journal of Medicine, reported an increased myocardial infarction risk associated with rosiglitazone across a pooled set of trials. The FDA restricted rosiglitazone prescribing in 2010 and partially lifted those restrictions in 2013 after re-analysis of a large trial did not confirm a statistically significant increase in MI. The exact odds ratio and confidence interval from the 2007 meta-analysis are commonly quoted in secondary sources, but this draft does not repeat a specific number tied to an unverifiable link; the original NEJM paper should be consulted directly for that figure.

Pioglitazone, approved the same year as rosiglitazone, did not carry the same cardiovascular signal in its own outcome trial and remains in use today, including in patients with non-alcoholic steatohepatitis, where trial evidence has shown histological benefit.

Incretin-based drugs: DPP-4 inhibitors and early GLP-1 agonists (2000s)

The "incretin effect," the observation that oral glucose triggers more insulin release than the same glucose given intravenously, was described decades before it became druggable. Native GLP-1 has a half-life of only one to two minutes, which made it clinically useless until chemists found ways to extend its action or block its degradation.

DPP-4 inhibitors (sitagliptin, first approved in 2006, followed by others through 2013) block the enzyme that degrades native GLP-1. They produce modest HbA1c reductions with very low hypoglycemia risk and weight neutrality, but two large cardiovascular outcome trials in this class reported an unexpected increase in heart failure hospitalization with one DPP-4 inhibitor, which is why current guidance cautions against this class in patients with existing heart failure.

Exenatide, derived from a peptide found in Gila monster venom, was approved by the FDA in 2005 as the first injectable GLP-1 receptor agonist, followed by a once-weekly formulation in 2012. It established the injectable incretin-mimetic mechanism that later, more potent GLP-1 drugs would build on.

GLP-1 receptor agonists: where cardiovascular outcome data changed the guidelines (2010s-present)

After the rosiglitazone controversy, the FDA required cardiovascular outcome trials for new diabetes drugs starting in 2008. That single regulatory decision produced the largest cardiovascular trial dataset in the history of diabetes pharmacotherapy, and it is the direct reason GLP-1 receptor agonists and SGLT2 inhibitors now carry cardiovascular and kidney indications that older drug classes never generated evidence for.

Liraglutide (LEADER trial) and semaglutide (SUSTAIN-6, and later SELECT in people with obesity but without diabetes) both demonstrated reductions in major adverse cardiovascular events compared with placebo in large randomized trials enrolling patients with established cardiovascular disease or high cardiovascular risk. Oral semaglutide, approved by the FDA in 2019, became the first oral GLP-1 receptor agonist on any market and showed cardiovascular non-inferiority in its own outcome trial. This article does not restate the specific hazard ratios and confidence intervals from these trials because the underlying citations in the prior version of this page could not be verified; anyone relying on exact effect sizes for a clinical decision should pull the primary trial publications directly.

SGLT2 inhibitors: glucose-lowering that turned out to protect the kidney and heart (2013-present)

SGLT2 inhibitors block glucose reabsorption in the kidney's proximal tubule, causing the kidney to excrete excess glucose in urine. The mechanism does not depend on insulin secretion or beta-cell reserve, which distinguishes it from every earlier oral drug class.

Canagliflozin (2013), dapagliflozin (2014), and empagliflozin (2014) reached the US market in close succession. Empagliflozin's cardiovascular outcome trial, published in the New England Journal of Medicine in 2015, is widely credited with shifting the entire guideline hierarchy toward risk-stratified prescribing, reporting reductions in cardiovascular death and heart failure hospitalization in patients with established cardiovascular disease. Canagliflozin's kidney outcome trial was stopped early for benefit in patients with diabetic kidney disease, and a related dapagliflozin kidney trial extended similar protection to people without diabetes at all, which is part of why SGLT2 inhibitors are now recommended even in patients whose glucose control is already adequate, if they have heart failure or chronic kidney disease. Current ADA guidance reflects this comorbidity-first logic, detailed in the current Standards of Care (2024 edition; confirm against the most recent annual update, since ADA standards are revised yearly).

Dual and triple incretin agonism: tirzepatide and beyond (2022-present)

Tirzepatide, approved by the FDA in 2022 for type 2 diabetes, activates both GLP-1 and GIP receptors. Its dual mechanism produced larger HbA1c reductions and greater weight loss than semaglutide in head-to-head trial data, and separately produced substantial weight loss in a large obesity trial in people without diabetes. This is the newest mechanism in the class progression described here, and its long-term cardiovascular and kidney outcome data are still less mature than the datasets available for the earlier GLP-1 agonists and SGLT2 inhibitors that already have completed dedicated outcome trials.

Insulin's changing role in type 2 diabetes

Insulin has shifted from a last-resort rescue therapy to a more precisely timed tool. NPH insulin (1950) and later basal analogs, insulin glargine (2000) and detemir (2005), gave clinicians flatter, more predictable coverage than earlier formulations, reducing nocturnal hypoglycemia. Combination pens pairing basal insulin with a GLP-1 receptor agonist, approved around 2016 to 2017, reduced injection burden and offset some of the weight gain historically associated with insulin intensification. Insulin remains an appropriate and sometimes necessary therapy at any stage of type 2 diabetes when oral and injectable non-insulin options do not achieve adequate control, and is not a marker of treatment failure.

Bariatric and metabolic surgery

A randomized trial comparing bariatric surgery to intensive medical therapy in patients with type 2 diabetes and obesity, published in the New England Journal of Medicine, showed that a meaningful minority of surgical patients achieved medication-free glycemic control at five years, a result that changed ADA guidance to include metabolic surgery as an option for patients with a BMI above 30 kg/m2 who do not reach adequate control with lifestyle and medication.

How diabetes classification itself has changed alongside treatment

A history of treatment is incomplete without noting that the definition of "type 2 diabetes" has not stayed fixed. Some patients present with an acute ketosis-prone episode that looks like type 1 diabetes at onset but later demonstrate preserved beta-cell function and can come off insulin, a phenotype that has been studied for roughly three decades and is still being refined in the clinical literature, as summarized in a recent review of ketosis-prone type 2 diabetes. Separately, malnutrition-related diabetes, a classification WHO once recognized and later removed, has been revisited in recent literature as a distinct entity now sometimes labeled type 5 diabetes, discussed in a 2026 paper on this reclassification. Neither of these subtypes changes standard type 2 diabetes management for most readers, but they are a reminder that "type 2 diabetes" is a working clinical category, not a single fixed biological entity, and that a patient whose presentation is atypical (ketoacidosis, rapid weight loss, onset in a setting of malnutrition) deserves specific evaluation rather than default type 2 treatment. Diabetic ketoacidosis of any cause is a medical emergency requiring urgent evaluation, not outpatient management.

What is established, what is plausible, and what is not established

Established by randomized trial evidence and reflected in current guidelines: metformin's safety and mortality-related benefit in overweight patients with type 2 diabetes; GLP-1 receptor agonists and SGLT2 inhibitors reducing major cardiovascular events and, for SGLT2 inhibitors, slowing kidney disease progression, in patients with established cardiovascular or kidney disease; rosiglitazone's cardiovascular signal being significant enough to prompt FDA restriction, later partially reversed; metabolic surgery producing durable glycemic remission in a subset of patients with obesity and type 2 diabetes.

Plausible but not fully settled: whether tirzepatide's larger short-term glycemic and weight effects translate into cardiovascular and kidney outcome benefits equivalent to or exceeding the older GLP-1 agonists and SGLT2 inhibitors, since its dedicated cardiovascular outcome trial data are not yet as mature as those for liraglutide, semaglutide, empagliflozin, or canagliflozin.

Not established from the material available to this review: exact numeric effect sizes (hazard ratios, percentage risk reductions, confidence intervals) for several landmark trials referenced in older versions of this article, because the specific citations attached to those numbers could not be verified against the primary literature during this review. Readers and clinicians who need those numbers for a decision should pull the primary trial publication rather than rely on a secondary summary, including this one.

A framework for evaluating a historical claim about a diabetes drug

Readers, students, and clinicians who encounter a specific number or claim about a diabetes drug's history (in this article, in a review article, or in marketing material) can use this sequence before treating the number as settled:

This does not replace a clinician's judgment or a pharmacist's label check. It is a way to keep a historical fact from being mistaken for current, individualized guidance.

References

• American Diabetes Association Professional Practice Committee. Standards of Medical Care in Diabetes 2024. Diabetes Care. https://diabetesjournals.org/care/article/47/Supplement_1/S1/153954/
• American Diabetes Association Professional Practice Committee. Introduction and Methodology, Standards of Care in Diabetes 2024. https://diabetesjournals.org/care/article/47/Supplement_1/S1/153954/Introduction-and-Methodology-Standards-of-Care-in
• Ketosis-Prone Type 2 Diabetes Mellitus: 3 Decades of Clinical, Pathophysiologic, and Therapeutic Insights (2026). https://pubmed.ncbi.nlm.nih.gov/42002040/
• Type 5 diabetes mellitus: revisiting malnutrition-related diabetes (2026). https://pubmed.ncbi.nlm.nih.gov/42635143/

Numeric effect sizes for UKPDS, EMPA-REG OUTCOME, LEADER, SUSTAIN-6, CREDENCE, and the 2007 rosiglitazone meta-analysis were removed or generalized in this revision because the specific citations attached to them in the prior draft could not be verified. An editor with primary-literature access should restore verified figures with correct citations before this page is published.