Letters

Glycaemia and vascular effects of type 2 diabetes

BMJ 2001; 322 doi: https://doi.org/10.1136/bmj.322.7296.1245/b (Published 19 May 2001) Cite this as: BMJ 2001;322:1245

Lowering glucose concentrations may not be of any value in itself

  1. Brian Budenholzer, director (brbudenh{at}ghnw.ghc.org)
  1. Clinical Enhancement and Development Group Health Cooperative, Network Services Division PO Box 204, Spokane, WA 99210-0204, USA
  2. University of Manchester Medical School and Royal Infirmary, Manchester M13 9PT
  3. 45 Bishopsthorpe Road, London SE26 4PA
  4. Oxford Centre for Diabetes, Endocrinology and Metabolism University of Oxford, Oxford OX3 7LF

    EDITOR—Stratton et al have documented that as glycaemic exposure increases, diabetic complications increase.1 They conclude that treatment of hyperglycemia will have substantial benefit, a conclusion reiterated by Tuomilehto.2 Yet reduction of glycaemic exposure did not have such benefit in the UK prospective diabetes study (UKPDS) randomised trial. 3 4 The data by Stratton et al suggest that reducing mean haemoglobin A1C concentration by 1% would reduce diabetes related deaths by 21%. Intensive treatment of hyperglycaemia for 10 years in UKPDS reduced haemoglobin A1C by nearly 1% (from 7.9% to 7.0%) yet did not reduce diabetes related deaths significantly.

    The conventionally treated group, with greater glycaemic exposure, experienced diabetes related death at a rate of 11.5 deaths per 1000 person years. On the basis of the data by Stratton et al, the intensively treated group should have experienced diabetes related death at a rate of 9.0 deaths per 1000 person years. Intensive treatment was, however, associated with only a non-significant decrease in diabetes related mortality.4 Similarly, the data by Stratton et al suggest that intensive treatment would result in significant reductions in adverse outcomes that include all cause mortality, stroke, myocardial infarction, and amputation. Reducing haemoglobin A1C by nearly 1% in the UKPDS, however, was not associated with significant reductions in any of these adverse outcomes.

    Treatment that significantly improves glycaemic control therefore does not achieve the predicted benefit. Does this mean that greater glycaemic exposure is a marker for adverse outcomes but not a cause? This would imply that the higher the haemoglobin A1C concentration the more attention needs to be paid to non-glycaemic treatment of diabetic patients, such as controlling blood pressure. Or does it mean that the treatments currently available to lower glucose harm diabetic patients as much as the lowering of blood glucose helps them? McCormack and Greenhalgh may be correct when they say that treatment with metformin improves outcomes in diabetic patients, not necessarily resulting from its glucose lowering effect, but that lowering glucose concentrations in itself is of little to no value in type 2 diabetes.3

    Footnotes

    • Competing interests None declared.

    References

    1. 1.
    2. 2.
    3. 3.
    4. 4.

    UKPDS is not a cohort study and analysis is misleading

    1. J K Cruickshank, senior lecturer in clinical epidemiology and consultant physician (diabetes/hypertension)
    1. Clinical Enhancement and Development Group Health Cooperative, Network Services Division PO Box 204, Spokane, WA 99210-0204, USA
    2. University of Manchester Medical School and Royal Infirmary, Manchester M13 9PT
    3. 45 Bishopsthorpe Road, London SE26 4PA
    4. Oxford Centre for Diabetes, Endocrinology and Metabolism University of Oxford, Oxford OX3 7LF

      EDITOR—In their two papers the UK prospective diabetes study (UKPDS) investigators potentially render their analyses invalid. 1 2 UKPDS was subrandomised into many smaller comparisons with limited sample sizes in the subgroups. The trial provided unique data on treatment outcomes for type 2 diabetes where there had previously been virtually no evidence. It cannot be repackaged as an observational or epidemiological study by ignoring its design to test treatments.

      In the paper by Stratton et al, the major treatments are hardly mentioned and not accounted for in the main analyses.1 If Stratton et al are conceding that the main glycaemic interventions were ineffective, both on large vessel disease and mortality, and of limited impact on microvascular outcomes if the hypertension factorial component is excluded, then perhaps the pooling of intensive and conventional treatments would be legitimate. That concession seems unlikely, although that is the major result from a meta-analysis.3 So a multiple treatment term would at the very least be required in the observational analysis. Was this tried or was it not significant? Similarly, interaction terms (for example, the effective metformin-obesity arm) would be needed. In most cohort studies, subjects are only included when clinically free of the end points of interest. But in UKPDS over 30% of patients already had target organ signs or damage. What about the impact of weight gain in the analysis (some 2.5-4 kg after taking sulphonylureas and 7 kg after taking intensive insulin over the median 10 years of the trial)?

      A comparison of impacts on absolute rather than relative risks in the main trial arms may be useful (table), because detailed numbers needed for treatment to benefit or harm can be calculated. These show the slight absolute impact on any end point achieved by intensive glycaemic interventions with currently available treatment. They have not been discussed by the UKPDS trialists, in promoting results of intensive over merely good glycaemic control that have been packaged as very good news for type 2 diabetes patients but are of limited, if any, added benefit.

      Comparison of significant absolute risk reduction (ARR) and numbers needed to treat (NNT, or reciprocal of absolute risk reduction) for different arms of UK prospective diabetes study (UKPDS)

      View this table:

      In the paper by Adler et al on the impact of systolic blood pressure, antihypertensive treatment was effective in that arm of the trial. A treatment term is adjusted for in the analysis. The complex factorial design where only those with blood pressures ≥160 or 90 mm Hg were eligible for that arm of treatment renders such adjustment too simplistic, probably underestimating the impact of blood pressure, and illegitimate because of allocation to several treatments. The only appropriate analysis would be of the patients not randomised to the hypertension “tight control” arm.

      Both these papers present erroneous data, which might have been filtered at the refereeing stage. Withdrawal might be considered.

      Footnotes

      • Competing interests None declared.

      References

      1. 1.
      2. 2.
      3. 3.

      Relation between diabetes and hyperglycaemia and cardiovascular disease has not been resolved

      1. R J Jarrett, emeritus professor of clinical epidemiology, University of London
      1. Clinical Enhancement and Development Group Health Cooperative, Network Services Division PO Box 204, Spokane, WA 99210-0204, USA
      2. University of Manchester Medical School and Royal Infirmary, Manchester M13 9PT
      3. 45 Bishopsthorpe Road, London SE26 4PA
      4. Oxford Centre for Diabetes, Endocrinology and Metabolism University of Oxford, Oxford OX3 7LF

        EDITOR—Stratton et al show a relation between glycaemia (measured as glycated haemoglobin) and mortality and various morbid events in patients with type 2 diabetes.1 The statistical independence of this relation is, however, open to doubt. In multivariate analyses the precision of a variable can affect the outcome. In the UK prospective diabetes study (UKPDS) report, glycaemia is the average of multiple measurements—and is therefore a comparatively precise estimate of the individual's state—while potentially confounding variables are, for the most part, single measures at baseline. In the accompanying paper, by contrast, it is blood pressure that is the precise measure.2 Incidentally, given that metformin treatment in the UKPDS appeared to have an effect on cardiovascular end points independent of its effect on blood glucose concentration, should it not have been included as a potentially confounding variable?

        Another problem in the interpretation of the results is the inclusion of the lowest glycaemic group which, for the duration of the trial (or up to an event) had average haemoglobin A1c values below 6%, below the upper limit of the authors' normal range (6.2%). A substantial proportion of the individuals in this group probably did not have diabetes as defined by the current guidelines of the World Health Organization or the American Diabetes Association. The fasting plasma glucose criterion used by the UKPDS was >6.0 mmol/l, 1 mmol/l less than that recommended by the WHO and American Diabetes Association. Statistically the point is important, for the lowest glycaemic group contributed the second largest number of person years to the analyses.

        The debate on the nature of the relation between diabetes and hyperglycaemia, and cardiovascular disease has a long history, and the UKPDS has not resolved it. Perhaps however, given the substantial beneficial effects of statins and angiotensin converting enzyme inhibitors in secondary prevention and of hypotensive drugs in primary prevention, the debate has become academic. Clinicians will continue to try to control glycaemia to prevent microvascular disease.

        Footnotes

        • Competing interests None declared.

        References

        1. 1.
        2. 2.

        Authors' reply

        1. Irene M Stratton, senior statistician,
        2. Carole A Cull, senior statistician,
        3. Susan E Manley, biochemist,
        4. Amanda I Adler, epidemiologist,
        5. H Andrew W Neil, university lecturer in clinical epidemiology,
        6. David R Matthews, consultant diabetologist,
        7. Rury R Holman, director
        1. Clinical Enhancement and Development Group Health Cooperative, Network Services Division PO Box 204, Spokane, WA 99210-0204, USA
        2. University of Manchester Medical School and Royal Infirmary, Manchester M13 9PT
        3. 45 Bishopsthorpe Road, London SE26 4PA
        4. Oxford Centre for Diabetes, Endocrinology and Metabolism University of Oxford, Oxford OX3 7LF

          EDITOR—Budenholzer's suggestion that reduction of glycaemic exposure was not beneficial in the UK prospective diabetes study (UKPDS) is incorrect. Intensive blood glucose control by either sulphonylurea or insulin substantially decreases the risk of microvascular but not macrovascular disease.1 The risk reduction for myocardial infarction (16% for a 0.9% haemoglobin A1c difference; P=0.052) was, however, entirely consistent with that seen in the epidemiological analysis (14% for a 1% haemoglobin A1c decrement). The improved risk reductions seen with metformin, above those expected from the glycaemic improvement achieved, have been discussed.2 Moreover, we showed in table 3 that the point estimates and 95% confidence intervals for the risk reductions obtained in the clinical trial were not inconsistent with the estimates obtained from the observational analyses of either baseline or updated mean haemoglobin A1c values. For example, the observational analysis of updated mean haemoglobin A1c concentrations for any diabetes related death showed a risk reduction of 21% (95% confidence interval 15% to 27%) for a 1% decrement in haemoglobin A1c, consistent with the risk reduction achieved in the clinical trial of 10% (−11% to 27%) for a 0.9% absolute difference in median haemoglobin A1c concentration.

          Cruickshank's concerns about the possible effects of treatment allocation on the relation of glycaemia and blood pressure to diabetic complications have already been addressed by us in detail. In our first paper we reported that in these models treatment of blood glucose itself had no association with any complication beyond that of mean updated haemoglobin A1c. In the second we showed that treatment of blood pressure had an effect over and above the updated mean systolic blood pressure for stroke, heart failure, and diabetes related deaths. Cruickshank suggests that we consider absolute rather than relative risks and numbers needed for treatment to benefit or harm, which we have already done.14

          When allocation to metformin is included in the model restricted to overweight patients, a significant metformin effect is seen on updated mean haemoglobin A1c only for the any diabetes related aggregate end point (P=0.044). Reanalysis of the data, following Jarrett's suggestion, to include updated high density lipoprotein cholesterol and low density lipoprotein cholesterol, together with both updated haemoglobin A1c concentrations and systolic blood pressure, shows the risk relations with glycaemia and blood pressure to be unchanged. The choice of reference category does not influence the gradient of the risk relations that were calculated from continuous data. The large number of people with updated mean haemoglobin A1c values <6%, however, allows us to show the continuity of risk over a wide range of glycaemia.

          Footnotes

          • Competing interests AIA has received fees for speaking from Bristol-Myers Squibb, SmithKline Beecham, and Pfizer. IMS has received support for attending conferences from Zeneca and Hoechst and fees for speaking from Hoechst. CAC has received support for attending conferences from Bristol-Myers Squibb, Novo Nordisk, and Pfizer and fees for speaking from Bristol-Myers Squibb and Novo Nordisk. DRM has received fees for speaking from Bristol-Myers Squibb, Novo Nordisk, SmithKline Beecham, and Lilly and research funding from Lilly. SEM has received support for attending conferences from Bayer and Novo Nordisk. RRH has received fees for consulting from Bayer, Boehringer Mannheim, Bristol-Myers Squibb, Hoechst, Lilly, Novo Nordisk, Pfizer, and SmithKline Beecham; support for attending conferences from Bayer, Bristol-Myers Squibb, Hoechst, Lilly, Lipha, Novo Nordisk, and SmithKline Beecham; and research funding from Bayer, Bristol-Myers Squibb, Lilly, Lipha, and Novo Nordisk.

          References

          1. 1.
          2. 2.
          3. 3.
          4. 4.
          View Abstract

          Log in

          Log in through your institution

          Subscribe

          * For online subscription