Shannon.doc

A Review of some meta-analytic comparisons of pharmacological
interventions for prevention of cardiovascular disease in Type 2
diabetes
A.G. Shannon
Warrane College, The University of New South Wales, & KvB Institute of Technology, North Sydney, NSW 2060, Australia This paper was written to mark the retirement of Professor Soonki Kim, friend and colleague

Introduction
There is a lack of consistent evidence in the literature about the prevention and treatment of
macro-vascular disease in Type 2 diabetes, this being one of the many potential complications
associated with having the disease. Intuitively one knows that blood pressure and cholesterol
levels have to be controlled, but how best to do this, either in general or for particular classes
of patients?
Meta-analyses of the literature have been carried out but even they do not convey an
unequivocal picture of appropriate procedures. This paper is an attempt to get an overall
perspective of what works in general, based upon studies considered by other authors in
previous meta-analyses. Follow up is required then for specific treatments as detailed in the
references.
We note that in some of the studies cited in previous meta-analyses:
• the methods of defining patients with diabetes mellitus (DM) varied; • some studies mixed up Type 1 and Type 2 in the one study; • some studies were confined to one gender; • some studies total numbers were not reported. Data from such studies were not included.
Given the above caveats it was felt appropriate and beneficial for discussion to take a broader
overview of the results in the research reports cited.
Meta-analysis
Meta-analysis is a statistical approach to aggregate and analyse summary statistics from a
number of studies (Glass et al, 1981). It is especially useful where studies disagree with regard
to the magnitude or direction of an effect. For instance, we have used the approach to compare
glycaemic control with human and porcine insulins by means of data on glycosylated
haemoglobin, fasting blood-glucose and mean blood-glucose levels in various reported studies
(Shannon et al, 1990).
A meta-analysis is much more structured and replicable than an ordinary narrative literature
review. Based on Chalmers and Lau (1993) we have developed a ten step procedure for
conducting meta-analyses:
• development of a protocol for conducting the meta-analysis; • identification of sources of information used; • definition of the criteria for the selection of trials for inclusion; • reading, classification, coding, scoring, evaluating and choosing of literature; • adjudication of differences among readers on the qualitative criteria; • development of questions, procedures, and analyses to pose of trials for inclusion; • reading of papers and answering of questions on the checklists; • combination of results and quality assurance of the data; • analysis, interpretation and reporting of results.
Type 2 Diabetes
One of the difficulties in the development of general guidelines for the treatment of Type 2
diabetes (and the need for this research) is that, unlike Type 1 diabetes, it develops gradually
rather than suddenly; that is, we know when it is diagnosed but not necessarily when it
commences. Thus, baseline data are ambiguous.
Figure 1 schematically represents the progress of Type 2 diabetes for those people who have
the right ‘cocktail’ of genes to be predisposed to Type 2 diabetes. (The nature versus nurture
debate is as alive here as it is in educational research. For instance, there is evidence, albeit
circumstantial, that that the diet of the pregna nt mother, particularly during the last six to eight
weeks of pregnancy when the adipose and islet cells are formed, can also play a role in the
subsequent development of NIDDM in the offspring (Shannon et al, 1997).)
Figure 1: The Progress of Type 2 Diabetes
Some people who are genetically predisposed to Type 2 diabetes become insulin resistant,
fewer still develop impaired glucose tolerance (IGT), and then a few move on to get Type 2
diabetes (or non-insulin dependent diabetes mellitus, NIDDM). If insulin resistant people
adopt appropriate lifestyles of diet and exercise, then the onset of NIDDM may be delayed and
its complications, such as cardiovascular disease, avoided or, at least, minimised.


Methods
The data here ha ve been combined by means of the following criteria:
• data are not that accurate in the meta-analyses when combined so we have used integer values since decimal precision adds nothing to the picture; • there are high rates of discontinuation in some studies; • the effect of aspirin on the primary prevention of cardiovascular disease in people with diabetes was not included because some studies included both types of diabetes and others did not report the original diagnostic criteria.
By using weighted averages and ranges (weighted according to original sample size) we have
a crude indicator of possible significant differences (but no measure of the extent of
significance). Thus the counter-intuitive result for a slight increase in stroke events with the
relatively greater intensive treatment in Table 1 is put into perspective when the ranges are
considered.
Results
With this very conservative approach the “significant” differences are with those marked with
‘*’ where the ranges do not overlap. The results are events per 1,000 person-years.
Table 1: Glycaemic Control (UGDP, VACSDM, UK PDS) Table 2: Blood Pressure Lowering (SHEP, Syst.Euro, HOPE) Table 3: Lipid Lowering Statins (4S, CARE, LIPID, VAHITS) Table 4: Lipid Lowering Fibrates (VA-HIT, DAIS, SENDCAP HHS) In so far as one can generalize in the absence of random sampling in the original studies, it can be observed that with these weighted averages: • intensive treatment for glycaemic control does not have a major effect on • anti-hyperintensive agents have significant benefits in reduction of the primary end • lipid lowering statins, but not fibrates, have significant effects in reducing strokes and marked effects in reducing coronary heart disease. This is not to imply that there are no other benefits in these strategies but merely to confirm that their effects are not consistent in the limited number of relevant studies available. Furthermore, the above global approach obscures the subtleties which may be observed with particular agents in each category. Huang et al (2001) also conducted meta-analyses. They searched MEDLINE (1966 to 2000) to identify randomized controlled trials in Type 2 diabetes which compared intensive medication control of risk factor levels in standard therapy or placebo. Some of their studies were the same as those included in Briganti et al (2000) and Colagiuri and Blest (2002). The results of our previous (above) approach accord well with those of Huang et al (except in two places) as can be seen in Table 5 in which the results are expressed as summary rate ratios. Table 5: Comparison with Huang et al
Conclusion
From the source studies we can say that blood pressure lowering and lipid lowering (statins)
produce definite reductions in the incidence of coronary heart disease and stroke, whereas
intensive glucose lowering reduced effects only slightly. A major problem is getting sufficient
statistical power because of the small number of studies which have Level I quality of
evidence.
The point of the foregoing analysis is to suggest that a global approach to the meta-analysis
can be useful in its simplicity by highlighting the important points in the big picture. To delve
further into those points is a separate issue for the diabetologist in the context of the treatment
of the individual patient,
As an encouragement to statisticians to get involved in these rapidly developing areas of
medical and social research, we conclude with the four point quality of evidence scale
recommended by the National Health and Medical Researc h Council of Australia after
adaptation from the United States Preventative Services Taskforce (Liddle et al, 1995:
Mitchell and Wang, 1996).
Table 6: Quality of Evidence rating Scale Controlled Trials
Epidemiological Evidence
A well-designed population based study or Well-designed case-control study, cohort
References
Briganti, Esther, M., O’Brien, Richard C. and McNeil, John J. Options for a National
Diabetes Vascular Prevention Strategy. Melbourne: Monash University, 2000.
Colagiuri, Stephen and Blest, James. Lipid-lowering therapy in people with type 2 diabetes.
Current Opinion in Lipidology. 2002, 13:617–623.
Glass, Gene V, McGaw, Barry & Smith, Mary Lee. Meta-analysis in Social Research.
Beverley Hills: Sage, 1981.
Huang, Elbert S., Meigs, James B. and Singer, Daniel E. The Effect of Interventions to
Prevent Cardiovascular Disease in Patients with Type 2 Diabetes Mellitus. The American
Journal of Medicine. 2001, 111: 633–642.
Liddle, Jeannine, Williamson, Margaret & Irwig, Les. Evidence Evaluation Tool. Sydney:
New South Wales Department of Health, 1995.
Mitchell, Paul & Wang, Jie Jin. Clinical Practice Guidelines for the Management of Diabetic
Retinopathy. Sydney: Department of Ophthalmology, University of Sydney, 1996.
Shannon, Anthony G, Colagiuri, Stephen & Miller, John J. Comparison of Glycaemic Control
with Human and Porcine Insulins – a Meta-analysis. Medical Journal of Australia. 1990; 152:
49.
Shannon, A G, Wong, C K & Wyndham J R. Gestational Diabetes Mellitus: A Meta-analysis.
Sydney: New South Wales Department of Health, 1997.
Abbreviations
BIP
Bezafibrate Infarction Prevention (study) Cholesterol and Recurrent Events (study) Diabetes Atherosclerosis Intervention Study Health Outcomes Prevention Evaluation Study Long-term Intervention with Pravastatin in Ischaemic Disease (study) St Mary’s, Ealing, Northwick Park Diabetes Cardiovascular Disease Prevention (study) Systolic Hypertension in the Elderly Trial United Kingdom Prospective Diabetes Study Veterans Affairs High-Density Lipoprotein Intervention Trial West of Scotland Coronary Prevention Study

Source: http://sunhome.pe.kr/tim/Papers/others/Shannon.pdf