Are Lower Fasting Plasma Glucose Levels at Diagnosis of Type 2 Diabetes Associated With Improved Outcomes?

U.K. Prospective Diabetes Study 61

  1. Stephen Colagiuri, FRACP1,
  2. Carole A. Cull, PHD2,
  3. Rury R. Holman, FRCP2 and
  4. For the UKPDS Group
  1. 1Diabetes Center, Prince of Wales Hospital, Randwick, Australia
  2. 2Diabetes Trials Unit, Oxford Center for Diabetes, Endocrinology and Metabolism, University of Oxford, Oxford, U.K


    OBJECTIVE—Type 2 diabetes may be present for several years before diagnosis, by which time many patients have already developed diabetic complications. Earlier detection and treatment may reduce this burden, but evidence to support this approach is lacking.

    RESEARCH DESIGN AND METHODS—Glycemic control and clinical and surrogate outcomes were compared for 5,088 of 5,102 U.K. Diabetes Prospective Study participants according to whether they had low (<140 mg/dl [<7.8 mmol/l]), intermediate (140 to <180 mg/dl [7.8 to <10.0 mmol/l]), or high (≥180 mg/dl [≥10 mmol/l]) fasting plasma glucose (FPG) levels at diagnosis. Individuals who presented with and without diabetic symptoms were also compared.

    RESULTS—Fewer people with FPG in the lowest category had retinopathy, abnormal biothesiometer measurements, or reported erectile dysfunction. The rate of increase in FPG and HbA1c during the study was identical in all three groups, although absolute differences persisted. Individuals in the low FPG group had a significantly reduced risk for each predefined clinical outcome except stroke, whereas those in the intermediate group had significantly reduced risk for each outcome except stroke and myocardial infarction. The low and intermediate FPG groups had a significantly reduced risk for progression of retinopathy, reduction in vibration sensory threshold, or development of microalbuminuria.

    CONCLUSIONS—People presenting with type 2 diabetes with lower initial glycemia who may be earlier in the course of their disease had fewer adverse clinical outcomes despite similar glycemic progression. Since most such people are asymptomatic at diagnosis, active case detection programs would be required to identify them.

    Diabetes is a serious and growing public health problem that results in reduced life expectancy and increased morbidity due to diabetes-specific complications such as blindness and amputation. The risk of coronary heart disease is two to four times greater in individuals with diabetes than in those without diabetes (1), and 52% of individuals with type 2 diabetes die from cardiovascular disease (2). People can have undetected type 2 diabetes for many years, with ∼50% of individuals with the condition being unaware of it (3). At the time of diagnosis, over half of patients have one or more diabetic complication (4). Microvascular complications at the time of diagnosis are common; Harris et al. (5) reported retinopathy rates of 21% at the time of diagnosis, and the U.K. Prospective Diabetes Study (UKPDS) reported rates of 37% (6). Since the development of retinopathy is related to the duration of diabetes, it has been estimated that type 2 diabetes may have its onset up to 12 years before clinical diagnosis (5).

    One possible strategy for reducing the type 2 diabetes burden is active case detection, with the rationale that diagnosis and intervention earlier in the disease process may more effectively prevent or delay the development of diabetes-related complications. This approach, however, has yet to be tested in a randomized controlled trial.

    We compared clinical outcomes in UKPDS participants according to their fasting plasma glucose (FPG) levels at presentation to determine whether a lower FPG at diagnosis, which may indicate an earlier time point in the natural history of the disease, is associated with improved clinical outcomes irrespective of randomized therapy for glycemia.


    Details of the 5,102 25- to 65-year-old patients newly diagnosed with type 2 diabetes who were recruited into the UKPDS (of the 7,616 referred) have been reported previously (7). In brief, 81% were self-reported Caucasian, 10% Asian Indian, and 9% Afro-Caribbean. The 2,514 patients not recruited were either unwilling to join the study or had one or more exclusion criteria, which included severe vascular disease, a myocardial infarction or stroke within the year preceding recruitment, or major systemic illness.

    After a 3-month dietary run-in period (low-fat, high-carbohydrate, high-fiber diet, with calorie restriction if obese), patients were stratified into three groups according to the mean of three FPG values taken on separate days. Individuals with hyperglycemic symptoms or mean FPG values >270 mg/dl (15 mmol/l) were termed “primary diet failure” and randomized to an intensive treatment policy primarily with sulfonylurea or insulin, with the additional possibility of metformin if overweight (>120% ideal body weight [Metropolitan life tables]). Asymptomatic patients with FPG values in the range of 108 mg/dl (6 mmol/l) to 270 mg/dl (15 mmol/l) were termed “main randomization.” These patients were randomly allocated either to a conventional treatment policy, primarily with diet alone, or to an intensive treatment policy as above. Patients with FPG values <108 mg/dl (6 mmol/l) were termed “diet satisfactory.” They were not randomized but maintained on diet alone unless the mean of three consecutive FPG values became ≥108 mg/dl (6 mmol/l), or they developed hyperglycemic symptoms, when they were randomized as described above.

    Patients were seen at three monthly intervals in a UKPDS clinic, with a detailed medical examination at entry and every 3 years that included retinal photography and asking men about erectile dysfunction. Retinal photographs were assessed using a modified Early Treatment of Diabetic Retinopathy Study (ETDRS) scale, and results were categorized using the worse eye/better eye system (8). Not all subjects had a retinal photograph taken at baseline. Interpretation of other clinical measures, such as electrocardiogram abnormalities, has been previously described (4).

    Biochemistry methodology has been reported previously (9). FPG values were measured in each center with central monitoring to maintain comparability; plasma creatinine levels were also measured locally. All other samples reported here were measured by the UKPDS central laboratory. Methods were upgraded during the study, and data were realigned to new methods after relevant laboratory comparisons (10). HbA1c was measured using high-performance liquid chromatography on a Bio-Rad Diamat Automated Analyzer with a reference range in nondiabetic subjects aged 25–65 years of 4.5–6.2%. Plasma lipid and lipoprotein were measured by standard laboratory methods (9).

    The predefined aggregate clinical end points (11) examined were any diabetes-related complications, diabetes-related deaths, all-cause mortality, myocardial infarction, stroke, peripheral vascular disease, and microvascular disease. Surrogate end points examined were a two-step progression of retinopathy on the ETDRS scale, peripheral sensory neuropathy (biothesiometer reading >25 V), the development of microalbuminuria (>50 mg/l) or proteinuria (>300 mg/l) in early morning urine samples, and a doubling of plasma creatinine.

    Statistical analyses were performed using SAS (12). Data are reported as the means ± SD, geometric mean (1-SD interval), median (interquartile range), or percentages. Patients were categorized into low, intermediate, and high levels of presenting FPG: <140 mg/dl (7.8 mmol/l) (World Health Organization 1985 criterion for FPG diagnostic of diabetes [13]), 140 to <180 mg/dl (7.8 to <10.0 mmol/l), and ≥180 mg/dl (10 mmol/l). The level of 180 mg/dl was chosen because this is the usual renal threshold for glucose and heralds the point at which patients may experience glycosuric symptoms leading to a diagnosis of diabetes. β-Cell function and insulin sensitivity were derived using homeostatic model assessment (HOMA) (14). Comparisons of continuous data used two-sample t tests, ANOVA, or, for nonnormally distributed data, the Wilcoxon’s signed-rank or Kruskal-Wallis test. χ2 tests were used for categorical variables with a Fisher’s exact test where cells contained <5% of the data. A Bonferroni adjustment was used to protect from type 1 error when performing multiple tests. Kaplan-Meier analysis was performed to compare survival times, with proportional hazards modeling to calculate relative risks with 95% CIs for aggregate end points. Relative risks for surrogate end points were calculated from frequency tables as Mantel-Haenszel estimates with 99% test-based CIs. Absolute risks were calculated as events per 1,000 patient-years. A two-tailed level of significance of P < 0.05 was used throughout. The analyses at each 3-year interval excluded those who had died, were lost to follow-up, or had no data available for a particular visit.


    Of the 5,102 subjects who participated in the UKPDS, data on the mode of presentation were available on 4,559. Of these, 2,446 (53.7%) presented with diabetic symptoms, 594 (13%) had an infection, and 96 (2.1%) had a diabetes-related complication. The remaining 1,423 (31.2%) were asymptomatic, with the diagnosis of diabetes being an incidental finding, e.g., detection of glycosuria on a routine medical examination. Proportions of Caucasian subjects were the same in the asymptomatic and symptomatic groups (84 vs. 84%), with fewer Afro-Caribbean subjects (5 vs. 7%) and more Indian Asian subjects (11 vs. 9%), respectively (χ2 test, P = 0.0036). Asymptomatic subjects were more often male (62 vs. 55%, P < 0.0001), had significantly less glycemia than the symptomatic group (FPG 184 mg/dl [10.2 mmol/l] vs. 218 mg/dl [12.1mmol/l]; HbA1c 8.1 vs. 9.6%; P < 0.0001 for both), and had greater mean body weight at presentation (83 vs. 81 kg, P < 0.0001). There were no significant differences between asymptomatic and symptomatic patients in the prevalence (38.2 vs. 36.3%) or severity (ETDRS grade 35 < 35 or worse: 7.4 vs. 6.6%) of retinopathy, abnormal biothesiometer measurement (>25 V: 4.5 vs. 5.2%), previous myocardial infarction (2.5 vs. 1.6%), previous cerebrovascular accident (1.2 vs. 1.3%), intermittent claudication (5.8 vs. 5.3%), or microalbuminuria (16.2 vs. 17.4%). However, people in the asymptomatic group had significantly less symptomatic neuropathy (3.9 vs. 5.9%, P = 0.0044) and erectile dysfunction (5.9 vs. 7.6%, P = 0.0046). No significant differences were seen between the two groups for any of the predefined UKPDS end points.

    Of the 5,088 UKPDS participants with FPG data available at presentation (Table 1), 65% had a high FPG value (≥180 mg/dl), 21.4% had an intermediate value (140 to <180 mg/dl), and 13.6% had a low value (<140 mg/dl), with corresponding median FPG values of 243 mg/dl (13.5 mmol/l), 160 mg/dl (8.9 mmol/l), and 126 mg/dl (7.0 mmol/l). Patients in the highest FPG group were older (P = 0.011), whereas more patients in the two lower FPG groups were male (P < 0.0001) or had gestational diabetes (P < 0.0001), and more patients in the low FPG group were current smokers (P = 0.029). Blood pressure and all lipid fractions were highest in the high FPG category (P < 0.0001 for all), and plasma creatinine values were marginally higher (P < 0.0001) in the two lower FPG categories. HOMA showed markedly greater β-cell function in the low FPG category compared with the intermediate and high FPG categories (80 vs. 54 vs. 23%; P < 0.00001). Although insulin sensitivity also differed significantly, the magnitude of the difference was small with no evidence of a trend (52 vs. 47 vs. 49%; P < 0.013).

    At presentation, fewer people in the low FPG group had any retinopathy compared with the intermediate and high FPG groups (33.3 vs. 35.7 and 39.2%, P < 0.0001), and fewer had severe retinopathy (7.6 vs. 10.0 and 16.3%, P < 0.0001). The low FPG group also had fewer patients with abnormal biothesiometer measurements >25 V (9.7 vs. 10.0 and 12.2%, P = 0.048) and fewer men with erectile dysfunction (4.4 vs. 5.8 and 8.1%, P = 0.047).

    After the run-in period, median HbA1c values remained significantly less in the low and intermediate FPG categories, with fewer patients entering the primary diet failure and main randomization stratifications. By the end of the study, fewer people in the low FPG group, compared with the intermediate or high groups, had progressed to increased treatment requirements (from diet to medication or from oral monotherapy to multiple therapy or using an insulin dose ≥0.7 units/kg (38 vs. 41 and 49%, P < 0.0001) (after adjustment for allocation). Although the median FPG and HbA1c remained lower over the study period in the group with initial FPG <140 mg/dl compared with the intermediate and high FPG groups, respectively, the rate of increase in median FPG over time in the three groups was identical (Fig. 1). Aligning the postrandomization FPG trajectories by left shifting the curves for the low and intermediate FPG group, such that they overlie the high FPG group curve, suggests that the intermediate and low FPG groups may possibly have been diagnosed ∼2 and 5 years earlier in their disease process.

    Table 2 shows the event rates and absolute and relative risks for the predefined clinical outcomes for patients in the low and intermediate FPG categories compared with the highest. Individuals in the lowest category had significantly lower risk for all complications except stroke. For individuals in the intermediate category, there was a significantly lower risk for all complications except stroke and myocardial infarction. The Kaplan-Meier plots for myocardial infarction (fatal and nonfatal) and microvascular disease are shown in Fig. 2.

    Two-step progression of retinopathy was significantly less at 3, 6, 9, and 12 years (P < 0.0001 for all) in the low and intermediate FPG groups compared with that in the high FPG group (Table 3). Similarly, the proportions of patients with abnormal biothesiometer readings were less in the low and intermediate FPG groups than in the high group at 3, 6, and 9 years but not at 12 years. Presence of microalbuminuria was recorded in fewer patients in the low and intermediate groups at presentation and after 3 years, but this was not observed thereafter. There were insufficient people with either proteinuria or doubling of plasma creatinine in the low FPG group to justify analysis of these variables.


    The UKPDS was not designed to study the impact of diagnosing diabetes at different times. It is known, however, that glycemic levels in type 2 diabetes increase progressively (15), suggesting that people with lower glucose levels at diagnosis may be at an earlier stage of the disease. We have attempted to address the question of whether earlier diagnosis of type 2 diabetes might be beneficial by analyzing outcomes for UKPDS participants according to their FPG level at presentation, irrespective of their allocated glucose control policy. People with lower FPG values at the time of diagnosis had fewer complications initially and fewer adverse clinical outcomes over time than people with higher FPG values, despite similar rates of increasing glycemia.

    Compared with individuals with high FPG values (≥180 mg/dl [≥10.0 mmol/l]), individuals with low FPG values (<140 mg/dl [<7.8 mmol/l]) had a significantly lower risk of any diabetes-related complications, diabetes-related death, all-cause mortality, myocardial infarction, peripheral vascular disease, and microvascular disease. Similarly, the intermediate FPG group (140 to <180 mg/dl [7.8 to <10.0 mmol/l]) had a significantly lower risk of any diabetes-related complications, diabetes-related death, all-cause mortality, peripheral vascular disease, and microvascular disease than the high FPG group. Compared with the intermediate FPG group, the low FPG group had a significantly lower risk of diabetes-related deaths and myocardial infarction. More caution needs to be exercised in interpreting results relating to peripheral vascular disease given the small number of events. The rate of rise in FPG levels after the institution of therapy in the study was virtually identical for the three FPG groups, although absolute differences persisted.

    If individuals with lower FPG values at diagnosis are representative of patients with diabetes detected at an earlier time point in the course of their disease, then left shifting their mean postrandomization FPG profiles, such that they are overlaid, suggests that the intermediate and low FPG groups may have been identified some 2–5 years earlier than individuals in the high FPG group. These estimates are similar to those seen in the Belfast Diet Study, in which FPG increased by 16 mg/dl (0.9 mmol/l) per year in type 2 diabetic subjects after the initiation of diet or oral therapy (16). These data suggest that procedures to identify people at an earlier stage could be beneficial in allowing therapeutic intervention before complications develop and in preventing or delaying their progression.

    Individuals diagnosed with lower FPG values may be representative of people with diabetes detected at an earlier time point in the course of their disease. Alternatively, they may have a different form of the disease with a lower susceptibility to complications. This hypothesis is supported by the fact that the mean age of diagnosis was not substantially different for the low, intermediate, and high FPG groups. In addition, the modifiable risk factors for diabetic complications tended to be more favorable in the low FPG group, apart from current smoking, and women in the low FPG group were more likely to have had gestational diabetes. However, the three groups had indistinguishable rates of increase in glycemia over time, suggesting that the natural history of diabetes in the groups was similar. The lower risk of adverse clinical outcomes observed in the lower FPG group could also reflect lead time bias. If this were the case, then we would expect the outcomes for all three FPG groups to be similar, provided that they were followed for a comparable duration of diabetes.

    An analysis of a mass-screening program based on urinary glucose levels, conducted in the former East Germany in the 1960s and 1970s, suggested that individuals found to have diabetes had an improved outcome than those presenting spontaneously with diabetes (17). However, there are no randomized controlled studies that have compared diabetes-related outcomes in screened and unscreened populations, and it seems unlikely that such a study will be performed in the foreseeable future because of logistic and ethical considerations (18). This lack of direct evidence means that only studies not primarily designed to address this question are available to provide some insight and information about the potential value of earlier diagnosis of type 2 diabetes. The UKPDS data presented here confirm that a lower FPG at the time of diagnosis of diabetes is associated with lower mortality and fewer macrovascular and microvascular outcomes. Further analyses suggest that the lower FPG values may reflect detection of individuals at an earlier time point in their disease process. Identifying people at a stage before they have developed diabetic complications gives greater scope for successful intervention and an opportunity to detect and treat concomitant hypertension and dyslipidemia. The better than expected reduction seen in the risk of microvascular disease suggests also that earlier treatment of hyperglycemia may have enhanced benefits. The lack of effect on myocardial infarction in the intermediate FPG group suggests that detection and intervention at yet lower FPG levels may be required to affect macrovascular as well as microvascular outcomes. Excess macrovascular risk, even at lower FPG levels, may be due in part to postprandial hyperglycemia (19).

    The presence or absence of diabetic symptoms at the time of diagnosis appears to be of little prognostic significance with regard to clinical outcomes. The differences observed between the asymptomatic and symptomatic groups with respect to sex, ethnicity, glycemia, and body weight may result from real differences in the development of diabetes between the groups, but could also reflect differences in the reasons for incidental testing. Thus, men may have been more likely to have had medical examinations for employment or life insurance reasons; hence, more of them were found at an asymptomatic stage. Similarly, cultural differences may affect the mode of presentation. The reductions in symptomatic neuropathy and reported erectile dysfunction do, however, reflect the slightly lower mean FPG values seen in individuals without symptoms. Actively seeking to detect people with diabetes who have lower FPG levels in the range associated here with improved outcomes is not easy because they will invariably be asymptomatic.

    Although this article suggests that some 65% of patients may be presenting late with FPG values >180 mg/dl (10 mmol/l), detecting asymptomatic individuals with undiagnosed diabetes is a major and potentially costly challenge. The justification for a screening program has been questioned because many people with diagnosed diabetes currently receive less than optimal care (18,20). Although the issue of availability of resources and services is legitimate, conclusive demonstration of improved outcomes with earlier detection of type 2 diabetes would provide strong support for the many diabetes organizations that currently advocate a diabetes screening program.


    Participating centers

    Radcliffe Infirmary, Oxford; Royal Infirmary, Aberdeen; General Hospital, Birmingham; St. George’s Hospital, London; Hammersmith Hospital, London; City Hospital, Belfast; North Staffordshire Royal Infirmary, Stoke-on-Trent; Royal Victoria Hospital, Belfast; St. Helier Hospital, Carshalton; Whittington Hospital, London; Norfolk and Norwich Hospital, Norwich; Lister Hospital, Stevenage; Ipswich Hospital, Ipswich; Ninewells Hospital, Dundee; Northampton General Hospital, Northampton; Torbay Hospital, Torbay; Peterborough General Hospital, Peterborough; Scarborough Hospital, Scarborough; Derbyshire Royal Infirmary, Derby; Manchester Royal Infirmary, Manchester; Hope Hospital, Salford; Leicester General Hospital, Leicester; and Royal Devon and Exeter Hospital, Exeter.

    Figure 1—

    Median FPG during the 3-month dietary run-in period and over the following 12 years, with patients classified according to whether they had low (<140 mg/dl [7.8 mmol/l]) (○), intermediate (140 to <180 mg/dl [7.8 to <10.0 mmol/l]) (•), or high (≥180 mg/dl [10.0mmol/l]) (black triangles) FPG levels at presentation. The lower panel shows the low and intermediate FPG groups left shifted by 3 and 5 years, respectively, such that their postrandomization trajectories are overlaid.

    Figure 2—

    Kaplan-Meier plots for myocardial infarction (fatal or nonfatal) and microvascular disease for patients classified according to whether they had low (<140 mg/dl [7.8 mmol/l]) (dotted line), intermediate (140 to <180 mg/dl [7.8 to <10.0 mmol/l]) (dashed line), or high (≥180 mg/dl [10.0 mmol/l]) (solid line) FPG levels at presentation.

    Table 1—

    Baseline characteristics according to FPG at presentation

    Table 2—

    Proportion of patients with predefined aggregate UKPDS clinical end points classified according to whether they had low (<140 mg/dl [<7.8 mmol/l]), intermediate (140 to <180 mg/dl [7.8 to <10.0 mmol/l]) or high (≥180 mg/dl [≥10.0 mmol/l]) FPG levels at presentation

    Table 3—

    Proportion of patients with progression of retinopathy, biothesiometer reading >25 V, and microalbuminuria at 0, 3, 6, 9, and 12 years after randomization to the UKPDS, classified according to whether they had low (<140 mg/dl [<7.8 mmol/l]), intermediate (40 to <180 mg/dl [7.8 to <10.0 mmol/l]), or high (≥180 mg/dl [≥10.0 mmol/l]) FPG levels at presentation


    The majority of grant support for this study came from the U.K. Medical Research Council; the British Diabetic Association; the U.K. Department of Health; the National Eye Institute; the National Institute of Digestive, Diabetes and Kidney Disease, National Institutes of Health; the British Heart Foundation; Novo-Nordisk; Bayer; Bristol Myers Squibb; Hoechst; Lilly; Lipha; and Farmitalia Carlo Erba.

    The cooperation of the patients and many National Health Services (NHS) and non-NHS staff at the centers is much appreciated. Other funding companies and agencies, the supervising committees, and all participating staff are listed in an earlier article (11).


    • Address correspondence and reprint requests to Prof. S. Colagiuri, Diabetes Center, Prince Wales Hospital, Randwick, 2031 NSW, Australia. E-mail: colagiuris{at}

      Received for publication 7 December 2001 and accepted in revised form 29 April 2002.

      A table elsewhere in this issue shows conventional and Système International (SI) units and conversion factors for many substances.


    | Table of Contents