What we learned from the UKPDS

Between its inception in 1977 and 2018 – the 20th anniversary of the publication of its primary findings – the UKPDS has generated an astonishing 93 original research papers. Here, we give an overview of all these papers, showcasing the vast breadth and depth of knowledge obtained from this seminal trial.

The primary findings: blood glucose and blood pressure

On 12th September 1998, the UKPDS investigators published five papers covering their primary findings. The two papers in The Lancet showed that intensive glycemic control with sulfonylureas or insulin significantly reduced the risk for microvascular outcomes, albeit with an increased hypoglycemic risk, but not macrovascular outcomes, and that metformin treatment in obese patients reduced the risk for combined micro- and macrovascular outcomes, as well as diabetes-related and all-cause mortality [1,2].

The three papers in The BMJ showed that controlling blood pressure in patients with type 2 diabetes resulted in a significant reduction in micro- and macrovascular outcomes and in mortality, irrespective of the specific medication used, and had a favorable cost–effectiveness ratio [3–5].

Click here to view and download summary slides of the primary and follow-up UKPDS findings.

And in 2008, the team published two papers in The New England Journal of Medicine revealing what had happened over the 10 years after the randomized treatment period, when patients resumed standard clinical care. This revealed a “legacy” effect of the period of tight glucose control, with patients formerly in the intensive group continuing to have a significantly reduced risk for microvascular complications, in addition to which the previously nonsignificant reduction in macrovascular outcomes attained significance [6]. There was no such effect for blood pressure, however [7].

Characterizing patients with newly diagnosed diabetes

The UKPDS enrolled 3867 patients, presenting the investigators with a unique opportunity to characterize a large number of people with newly diagnosed diabetes. They learned the following.

About demographics

• Participants were more likely to be male, but the female participants were more obese [8].
• Being a UK study, 83% of the participants were unsurprisingly of European ethnicity.
  • However, Asian and Afro-Caribbean participants were younger than the Europeans, and Afro-Caribbeans had the highest fasting plasma glucose levels and more severely impaired beta‐cell function, although they were more insulin-sensitive [9].

About their diet

• Participants on the whole consumed less fat relative to the general UK population, but more protein, with carbohydrate intake being on a par.
• Energy intake appeared to be reduced relative to the general population, but the investigators suspected this to be an effect of inaccurate self-reporting [10].

About hypertension

• Forty percent of men and 53% of women had hypertension, going by the then accepted threshold of 160/95 mmHg, but only 23% and 42%, respectively, had previously been diagnosed, and just 12% and 26% were taking antihypertensive medication [11].

About biochemical factors

• The majority of participants had markedly reduced insulin sensitivity and only moderately impaired beta-cell function, in line with the high prevalence of obesity, although a minority were lean with significant beta-cell impairment [12].
• The participants, particularly the female ones, had a poorer lipid profile than people without diabetes, characterized by high triglycerides and low levels of high-density lipoprotein cholesterol [13].
• They had elevated levels of albumin in the urine [14].
• Leptin levels did not differ by ethnicity, but correlated positively with fasting plasma insulin concentrations [15].

And their association with existing complications

• Existing micro- and macrovascular complications were not associated with each other and were individually associated with specific risk factors; for example, stroke with hypertension and retinopathy with hyperglycemia [16].
• Markers of renal damage were associated with the degree of hyperglycemia (in the case of urinary N-acetylglucosaminidase excretion) and hypertension (in the case of urinary albumin excretion) [17].

The progression of type 2 diabetes

The UKPDS participants underwent comprehensive assessments at baseline and every 3 years during the 10-year randomized phase of the study. The fact that the glucose and blood pressure targets were considerably more relaxed than they are nowadays gave the investigators insights into the natural history of type 2 diabetes:

• Diabetes progressed irrespective of efforts to control blood glucose, with the proportion of patients achieving control with their allocated monotherapy declining over time [18].
• Likewise, nearly half of patients taking sulfonylureas required an additional medication within 6 years of starting treatment [19].
• However, urinary albumin excretion barely changed over 3 years in the dietary intervention group, suggesting that diabetic nephropathy is a slowly evolving process [20].

The influence of baseline factors

• Ethnicity did not influence glycemic control during follow-up, but Caucasian patients put on the most weight while Afro-Caribbeans had the largest increases in systolic blood pressure but more favorable changes in lipid profiles [21].
• Age at diagnosis had variable effects on the likelihood of complications being present at baseline and occurring during follow-up. For example, myocardial infarction and neuropathy were more likely to be present in older patients at baseline, whereas retinopathy was not but occurred more frequently in older patients during follow-up [22].

The role of autoantibodies

Although type 2 diabetes is not perceived as an autoimmune process, a substantial proportion of patients do have islet autoantibodies, potentially identifying patients who actually have latent autoimmune diabetes in adults (LADA). Several UKPDS publications focused on the predictive power of autoantibodies and found the following:

• Autoantibodies to islet-cell cytoplasm and glutamic acid decarboxylase were most common in younger patients (21–34% of those ≤65 years) and these patients had a relatively low BMI and a high level of glycated hemoglobin (HbA1c), consistent with autoimmune-driven disease.
  • The majority of these patients became insulin-dependent within 6 years [23].
• Autoantibodies to protein tyrosine phosphatase IA-2 were very rare but highly predictive of insulin dependence [24].
• Autoantibody-positive patients initially responded well to intensive pharmacologic treatment, despite the rapid progression to insulin dependence [25].
• The SOX‐13 autoantigen did not distinguish between type 1 diabetes patients (from the Fremantle Diabetes Study) and UKPDS participants and is likely a “non‐specific marker of tissue damage associated with chronic hyperglycaemia” [26].

Control of blood glucose

The UKPDS used the only three classes of antidiabetes medication that were available at the time, in addition to dietary intervention, and the continued progression of diabetes in most patients meant that many ended up on combination therapy.

Specific therapies

• Three months of dietary intervention resulted in near normalization of blood glucose and corresponding improvements in lipid profiles, but only for 16% of patients, and this was related to energy restriction rather than to bodyweight [27,28].
• Acarbose significantly improved glycemic control over 3 years versus placebo, whether given alongside diet, metformin, insulin, a sulfonylurea, or a combination [29].
• Adding metformin to a sulfonylurea improved glycemic control during 3 years of treatment without resulting in more hypoglycemia [30].
• And the same was true when adding insulin to a sulfonylurea during 6 years of treatment [31].

Comparative efficacy over time

• After 1 year of follow-up, efficacy and hypoglycemic risk appeared similar for sulfonylureas and insulin, although both were associated with weight gain relative to diet alone [32].
• Changes in fasting plasma glucose correlated with changes in HbA1c, regardless of treatment allocation [33].
• The medications (including metformin) had comparable efficacy at the 3-, 6-, and 9-year reports, but only a minority of patients achieved glycemic control and by 6 years up to two-thirds of the patients had required intensification of therapy merely to achieve glucose levels below 15 mmol/L [34–36].
• Hypoglycemia rates were generally low, but were elevated in patients taking sulfonylureas and highest of all in those using insulin [37].

Genetic findings

Associations with diabetes subtypes

• Two papers strengthened the evidence that LADA and type 1 diabetes are closely related, with the HLA-DRB1 and HLA-DQB1 susceptibility genes and variation within the insulin gene region conferring susceptibility to both subtypes [38,39].
• In line with this, HLA-DRB1 and HLA-DQB1 were most common in younger UKPDS participants and were associated with the presence of islet autoantibodies [40].
• The butyrylcholinesterase gene K variant and variants in the sarco(endo)plasmic reticulum Ca²⁺-transport ATPase gene were associated with type 2 diabetes [41,42].
• But variants in the beta‐cell ATP‐sensitive K‐channel genes SUR1 and Kir6.2 were not linked to diabetes [43].
• MODY3 variants and a mutation in the tRNA^Leu(UUR) gene were very rare contributors to type 2 diabetes risk in the UKPDS population [44,45].
• The insulin receptor substrate-1 gene was associated with insulin resistance in people with diabetes, whereas the β3-adrenergic-receptor gene was associated with obesity [46].

Associations with complications

• Variants in the paraoxonase 2 gene had “at best, a small effect” on the risk for developing renal dysfunction [47].
• But one variant of the angiotensin-converting enzyme gene was associated with more severe microalbuminuria and another with an increased risk for myocardial infarction, with the effect strongest in people with low cholesterol levels [48,49].

The long-term consequences of diabetes

A vital factor underpinning the importance of the UKPDS was simply that no one had previously monitored such a large cohort of patients with newly diagnosed diabetes for so long, so seemingly simple questions like how many patients develop nephropathy over time were unanswered.

Incidence and progression of microvascular complications

• Evidence of retinopathy was already present in around a third of UKPDS participants at baseline, and was associated with higher blood glucose and blood pressure [50].
• The overall prevalence of nephropathy was around 7%, but this rose to 25% during 10 years of follow-up, with around 2–3% of patients annually developing the first signs of nephropathy or progressing to a more severe stage [51].

The effects of blood glucose and blood pressure control

The mission statement of the UKPDS was to examine the effect of blood glucose (and later blood pressure) control on the incidence of micro- and macrovascular complications of diabetes. This resulted in the landmark papers published in 1998 and 2008, but also in a large number of other publications looking at additional factors influencing patients’ risk for these outcomes. Two notable papers in The BMJ in 2000 [52,53] looked at the relationships between cumulative blood glucose and blood pressure exposure during follow-up and outcomes, irrespective of treatment allocation.

• There was a linear relationship between systolic blood pressure and outcomes, with higher pressure conferring greater risk for both microvascular outcomes and myocardial infarction (the most frequent macrovascular outcome).
  • The lowest risk was observed at blood pressures below 120 mmHg [52].
• The same was true for blood glucose, although the effect was stronger for microvascular outcomes, with the effect on myocardial infarction plateauing at an HbA1c of 9–10%.
  • The lowest risk was observed at HbA1c levels below 6% [53].
  • Baseline levels also played a part, with lower baseline glucose equating to lower risk during follow-up, despite a similar deterioration in glycemic control as seen for higher baseline glucose [54].
  • The protective effects against macrovascular outcomes, largely myocardial infarction, were confirmed in a meta-analysis also involving the ACCORD, ADVANCE, and VADT trials [55].
  • A more recent individual participant data meta-analysis involving the same trials confirmed the same for nephropathy and retinopathy [56].
• The effects of both approaches were additive, ie, patients receiving both intensive glucose control and intensive blood pressure control did better than those allocated to one or neither [57].
• Effects to achieve tight glucose or blood pressure did not impair patients’ quality of life, but developing microvascular complications or hypoglycemia did [58,59].

Intensive treatment and retinopathy

As the most frequent microvascular outcome, retinopathy was explored in more detail in two other papers.

• Blood pressure control markedly reduced the risk for all measures of retinopathy and its severity, up to and including blindness [60].
• The duration of hyperglycemia was as important as its severity for predicting the risk for photocoagulation [61].

Other influences on…

…microvascular outcomes

• The development of renal impairment and albuminuria were both associated with baseline systolic blood pressure, urinary albumin, plasma creatinine, and Indian-Asian ethnicity, but the individual endpoints were also associated with distinct sets of other risk factors, suggesting the two are “not linked inexorably” in diabetes patients [62].
• Besides blood pressure and glucose, not smoking was “somewhat counter-intuitively” associated with an increased risk of both the development and progression of retinopathy, and older age and male sex were associated with progression [63].
  • The need for photocoagulation was very rare – just 2.6% over 9 years of follow-up – although the risk was markedly higher among patients with retinopathy at baseline; baseline micro-aneurysms were highly predictive of progression during follow-up [64,65].

…peripheral vascular disease

• Pre-existing cardiovascular disease and current smoking were the strongest independent predictors of peripheral vascular disease, but older age, higher HbA1c, higher systolic blood pressure, and lower high-density lipoprotein cholesterol were also predictive [66].

…macrovascular outcomes

• Of the ethnic groups in the UKPDS, people of Afro-Caribbean ethnicity had the lowest risk for most vascular endpoints and Asian Indians had the highest risk over the long-term [67,68].
• For metabolic factors, neither hyperinsulinemia nor insulin sensitivity at baseline were significantly associated with macrovascular risk, but hyperglycemia was, as was the presence of the metabolic syndrome [69–71].
• Stroke risk was associated with baseline age, male sex, hypertension, and atrial fibrillation [72].
• Coronary heart disease was likewise associated with high blood pressure, and also with high levels of low-density lipoprotein cholesterol, triglycerides, fasting plasma glucose, and HbA1c. There was also a relationship with low levels of high-density lipoprotein cholesterol, and with a history of smoking [73].
• Subclinical myocardial infarction at baseline was associated with fatal myocardial infarction during follow-up, as were older age, higher blood pressure, higher urine albumin level, and higher HbA1c levels; the last factor was also associated with fatal (vs nonfatal) stroke [74,75].
• The predictive power of the cardiovascular risk marker homocysteine might be underestimated by half over 10 years, because of intra-individual variation in levels over time [76].

The predictive models that emerged

The wealth of data from the UKPDS allowed the investigators to develop predictive models to help guide the management of patients with type 2 diabetes. They released the following:

• The UKPDS Risk Engine, for predicting coronary heart disease and stroke [77,78].
• The UKPDS Outcomes Model: seven equations to predict cardiovascular disease, cerebrovascular disease, amputation, blindness, nephropathy, risk factor progression, and mortality [79].
  • The investigators validated the model using the post-trial follow-up data, and then further developed the model to include these follow-up data [80,81].
• The trial data were also used to produce a method to estimate life expectancy in people with type 2 diabetes [82].

The economics of diabetes management

The UKPDS investigators produced the following data relating to healthcare economics:

• Inpatient and non-inpatient costs for a range of diabetes complications including nonfatal and fatal stroke and myocardial infarction, heart failure, and blindness in one eye – calculated from the original randomized trial and then updated based on the post-trial monitoring data [83,84].
• The finding that intensive blood glucose control increased treatment costs, but also reduced the cost of complications and extended the time patients spent free of those complications [85].
• That the total costs of blood pressure treatment were less for the beta blocker atenolol than for the angiotensin-converting enzyme inhibitor captopril but the treatments were equally effective [86].
• The treatments used in the UKPDS to control blood glucose and blood pressure had acceptable costs per quality-adjusted life–year, with metformin actually reducing costs while extending life expectancy [87,88].
• Extrapolated to the population of England, the costs of intensively treating all diabetes patients with the trial medications were equivalent to “a small fraction of the NHS's spending plans” [89].

Other findings

And there were several other one-off papers, covering subjects ranging from pregnancy to the appropriate use of aspirin, that do not entirely fit into a theme. The findings are summarized below:

• The 16 women who became pregnant during the UKPDS had very poor outcomes, with just eight of the pregnancies leading to live births; there were five miscarriages, two terminations, and one intrauterine death [90].
• Most patients with pre-existing cardiovascular disease were taking aspirin, but two-thirds of those without were not, despite being at high cardiovascular risk [91].
• Less than half of the patients who died during the trial had diabetes noted on their death certificates [92].
• There was no evidence that improving or not improving people’s health affected the quality of neighborhood in which they resided [93].

By Eleanor McDermid

medwireNews is an independent medical news service provided by Springer Healthcare. © 2018 Springer Healthcare part of the Springer Nature group