In these observational studies there is a roughly linear relationship between CHD risk (plotted on a logarithmic scale) and blood cholesterol. This implies that the proportional reduction in CHD risk associated with a particular prolonged absolute cholesterol difference is similar throughout the range above about 3 mmol/l. So, for example, a prolonged difference of about 1 mmol/l in plasma cholesterol corresponds to about 50% less CHD, irrespective of the cholesterol level. Hence, the absolute size of the reduction in CHD produced by lowering cholesterol may be determined more by an individual's overall risk of CHD than by just their initial cholesterol level (Figure 3a), and may well be greatest in those who, as a consequence of their medical history (e.g. pre-existing vascular disease, diabetes, etc) or of some other factors, are at special risk of CHD (Figure 3b). Consequently, the eventual reductions in risk produced by the new drugs that can lower blood cholesterol substantially may well be worthwhile not only among certain "hypercholesterolaemic" individuals but also among those who, although considered to have "normal" UK cholesterol levels, are for some other reason at high risk.
Randomised trials are more relevant than observational studies to assessing how rapidly the CHD avoidance that is associated epidemiologically with a prolonged cholesterol difference can be achieved by treatments that lower blood cholesterol. In the randomised controlled trials of drugs or diets conducted so far, the average difference in cholesterol between treatment and control was only about 10%, and the mean treatment duration was only about 4 years. The observational studies suggest that a 10% difference in Western cholesterol levels that has persisted for decades eventually yields about a 30% difference in CHD. Overall, the results of these previous randomised controlled trials suggest that within just a few years of starting treatments which produce a 10% cholesterol reduction in middle age there is a reduction in CHD that is at least half as big as that from a long-term 10% difference in cholesterol (Table 1).
TABLE 1: CHD reduction by DURATION of 10% cholesterol difference: a systematic overview of 22 unconfounded randomised trials 8.
before CHD event
Difference in total CHD
from 10% cholesterol
(± 1 s.d.)
|14 "short" trials||1-4 years||1-2 years||9% ± 5|
|8 "longer" trials||5-7 years||3 years||22% ± 3|
|Observational studies||-||Decades||About 30%|
More prolonged treatment resulted in larger CHD reductions (Table 1), with the trials that lasted 5-7 years producing a highly significant 22% ± 3 CHD reduction 8. This large effect is particularly remarkable because at the time of CHD events in such trials the average duration of treatment was only about 3 years. Moreover, treatments that produced larger cholesterol reductions produced larger CHD reductions (Table 2), with significant reductions in CHD produced in patients both with and without a history of CHD (Table 3) 8,9. In these trials, there was a statistically significant reduction in fatal CHD (9% ± 3), and an even greater reduction in non-fatal MI (19% ± 4).
TABLE 2: CHD reduction by SIZE of cholesterol difference in trial 8
TABLE 3: Effects on CHD in primary and secondary prevention 8
These trials, together with the epidemiological evidence, suggest that a 1.5 mmol/l reduction in cholesterol in middle or old age might, within just a few years, reduce total CHD by about one-third and reduce fatal CHD by about one-quarter. This is supported by the small POSCH randomised trial of ileal surgery to reduce intestinal fat resorption among post-MI patients 10. In that study, a cholesterol reduction of about 1.5 mmol/l was maintained for about 10 years, and there was a significant 35%±10 reduction in total CHD (and a promising, though not significant, 28%±18 reduction in fatal CHD). Similar large reductions in cholesterol can now be conveniently achieved with drugs such as the HMG CoA reductase inhibitors (see below).
Some observational studies have found low cholesterol to be associated with increased rates of death from certain non-CHD causes (e.g. cancer, respiratory disease, trauma, haemorrhagic stroke 3). It is unclear, however, whether this inverse association is causal or due to some form of confounding (with certain diseases or habits causing both death and lower cholesterol 11). For example, preclinical cancer can itself reduce blood cholesterol (as can chronic hepatitis B virus infection) and, once any such effects are appropriately allowed for, a direct relationship between blood cholesterol and overall mortality extends down to well below 4 mmol/l 6,12.
Randomised evidence is not subject to such biases, but the previous trial results are still very much subject to the play of chance. Moreover, in the previous randomised trials (which included both primary and secondary prevention of MI), 40% of all deaths were from causes other than CHD. If a 10% cholesterol reduction does produce a 10% reduction in fatal CHD within a few years, then the expected reduction in total mortality would have been only about 6%. But, even in combination, the previous trials were too small for reliable detection of such an effect, especially since the non-cardiac deaths ran somewhat against treatment 8. This slight excess of non-cardiac deaths has attracted much comment over the years 13-16, but it could well be largely or wholly due to the play of chance 8,17. For, it is only marginally significant (P=0.02) when the primary and secondary prevention trials of drugs and of diets are, appropriately, taken together, and is the sum of a number of smaller statistically non-significant differences spread over several different diseases.
If lowering cholesterol really did produce any substantial adverse effect on non-CHD mortality then this should have been clearly evident in the POSCH study of ileal surgery 10, since in that study a much larger cholesterol reduction (1.5 mmol/l) was maintained for much longer (10 years) than in previous studies. But, although there was no excess whatever of non-fatal cancers or of deaths due to non-CHD causes, the POSCH study was small and so the trend towards fewer CHD deaths and a lower total mortality was not statistically significant.
Previous cholesterol-lowering drugs and diets did not produce large effects on blood cholesterol, either because some side-effect led to poor compliance or because they were not particularly effective. In contrast, the HMG CoA reductase inhibitors (such as simvastatin, lovastatin, pravastatin and fluvastatin) are well-tolerated and produce substantial lowering of plasma LDL cholesterol (1.5 mmol/l or more), along with a small increase in HDL cholesterol. These drugs all act by inhibiting 3-hydroxy-3-methyl glutaryl coenzyme A (HMG CoA) reductase, a rate-limiting enzyme in the hepatic synthesis of cholesterol 18. This reduces production by the liver of LDL precursors and, perhaps more importantly, increases the normal clearance by the liver of excess LDL cholesterol from the bloodstream.
More than a million patients worldwide are now being treated with HMG CoA reductase inhibitors. Clinical trial experience in thousands of patients suggests that, at least in the first few years, these cholesterol-lowering agents are effective and well-tolerated 19,20. Significant elevations in liver enzymes (alanine transaminase [ALT] and asparate transaminase [AST]) are produced in about 1% of patients but are usually reversible, with only a very few cases of clinical liver failure having been observed. Like other cholesterol-lowering drugs, reductase inhibitors can, in about 0.1% of patients, produce muscle pain or weakness associated with elevation of creatine kinase (CK). This also generally resolves within a few weeks of stopping treatment, and in only a very few cases has it progressed to significant myopathy. Concerns that decreased cholesterol synthesis might lead to substantial reduction in the availability of cholesterol for vital functions (such as the maintenance of cellular membrane structure and the synthesis of steroid hormones and bile acids) are not supported by the available human data 18,21. Careful and extensive human ophthalmological studies 20,22 have not reproduced earlier findings of lens opacities or other visual problems in beagle dogs given massive doses of reductase inhibitors.
Simvastatin has been shown to be about twice as effective (on a mg-per-mg basis) at lowering cholesterol as lovastatin and pravastatin 23,24. The simvastatin regimen proposed for the Heart Protection Study has been extensively tested in a randomised placebo-controlled pilot study of over 600 patients, with 3-year follow-up reported so far 20. High-risk patients were randomly allocated to receive 40 mg simvastatin daily, 20 mg simvastatin daily or matching placebo, and then carefully assessed for efficacy and side-effects at 2-6 monthly intervals in special nurse-run clinics (like those in the present study). Treatment and detailed follow-up is being maintained so that the pilot study can continue to give guidance (being always more than 4 years ahead) as to the longer-term safety and biochemical effects of the simvastatin regimen being studied in the Heart Protection Study.
Substantial and sustained effects on lipids: In the pilot study, simvastatin reduced total cholesterol by 28% (2.0 mmol/l) and LDL cholesterol by 40% at 8 weeks after randomisation. Compliance was good, and so these lipid differences were largely maintained at 3 years (e.g. 24%, or 1.7 mmol/l, reduction in total cholesterol). There was only a moderate difference between the cholesterol reductions produced by 40 mg daily simvastatin and by 20 mg daily simvastatin (26% versus 22% reduction respectively at 3 years). But, if the effect on CHD is approximately proportional to the effect on cholesterol 8-10 then the statistical power of a cholesterol-lowering study would be roughly proportional to the square of the cholesterol reduction. So, a study of 20 mg daily simvastatin might have to be about a third larger than one of 40 mg daily simvastatin to have equivalent statistical "power" (since the square of 26 is about one-third larger than the square of 22). Given that there was no difference in the incidence of any important blood abnormalities or possible side-effects between patients allocated 40 mg and those allocated 20 mg (or, indeed, those allocated placebo: see below), the present study uses 40 mg daily simvastatin.
Lack of evidence of any side-effects: During more than 3 years of follow-up in the pilot study, no difference at all was observed between the simvastatin and placebo groups in the responses to the regular general and specific enquiries about symptoms developing. Nor were there any differences between the simvastatin and placebo groups in the numbers of possible adverse drug reactions reported or in the small number of patients developing significantly elevated levels of liver enzymes (ALT >2x upper limit of normal [ULN]: 4 patients on 40 mg vs 6 on 20 mg vs 4 on placebo) or of creatine kinase (CK >2x ULN: 7 vs 9 vs 8 respectively). Concerns that reductase inhibitors might cause sleep or mood disturbances were addressed by standardised questionnaires completed 2-3 years after randomisation. These found no evidence of any effect on any aspects of sleep, anxiety, aggression, depression, etc. Similarly, extensive ophthalmologic examinations at 6 and 18 months after randomisation found no evidence of any effect of simvastatin on cataract or vision. This confirms other studies, including the large EXCEL study among 8000 patients of the closely related reductase inhibitor, lovastatin 19,22.
In summary, the reductase inhibitor regimen proposed for the present study has been extensively tested in a continuing pilot study of over 600 patients, and is currently being studied in the 4500 patient SSSS trial (see Table 4): 40 mg daily simvastatin is biochemically effective and appears to be remarkably free of serious side-effects, at least in the medium term. Long-term experience with reductase inhibitors is more limited, so future mortality studies must assess not only the efficacy but also any major long-term side-effects. The Heart Protection Study will do this particularly reliably (see data monitoring in Section 2.5).
Largely as a result of the lack of a clear mortality reduction in the previous trials, there is substantial uncertainty both in the medical profession and in the general population about any net benefits of the currently available methods of lowering cholesterol. Indeed, one common view is that use of them for reduction of blood cholesterol in middle or old age may well produce no significant improvement in survival, with a reduction in CHD deaths balanced by an increase in other deaths 13-16. Studies of intermediate outcome measures (i.e. of coronary atheroma regression, of non-fatal CHD or even of CHD death) are not able to address such concerns. Instead, trials are needed that can address directly the common medical question: "Will such treatment prolong my patients' lives or will it merely reduce their risk of dying from a heart attack but increase their risk of dying from something else?".
In principle it might appear preferable to use dietary means to lower cholesterol in such a trial. But, diets in long-term trials are likely to produce only moderate cholesterol reductions (perhaps 5-10% 25), so that impracticably large numbers of patients (perhaps some hundreds of thousands) would need to be studied. In contrast, the substantial cholesterol reductions now conveniently available with the HMG CoA reductase inhibitors provide an opportunity to assess directly the effects of this cholesterol-lowering therapy on total mortality in studies of practicable size.
Clear evidence from trials of these agents about their effects on total mortality and on non-CHD death would help establish more appropriate and cost-effective guidelines for the drug treatment of patients at high risk of CHD. Such trials will, however, still need to involve prolonged treatment, to be really large (i.e. some thousands of deaths), and to include a wide range of patients at substantial risk of death from CHD.
It has been suggested that the current randomised trials of HMG CoA reductase inhibitors can be relied on to provide adequate evidence about the effects of lowering blood cholesterol on total mortality 15,16. Those trials were, however, designed primarily to assess the effects of cholesterol lowering only on CHD. Consequently, principal investigators from each of the trials have publicly agreed that their studies have limited ability to detect the sort of effects on total mortality that it is realistic to hope for except perhaps in the special circumstance of patients who have already suffered an MI, where nearly all deaths are due to CHD 26. But, this special circumstance of "secondary" prevention includes only a fraction of the wide range of people who have a moderately elevated risk of CHD and for whom reliable evidence is needed about the efficacy and the safety of such cholesterol-lowering therapy (in particular, any real effects on non-CHD mortality and major morbidity).
In considering total mortality, suppose that in the current reductase inhibitor trials these cholesterol-lowering drugs have no material effect on non-CHD deaths and, as is suggested by the previous randomised trials of other treatments 8-10, that there is an approximately one-to-one relationship between the percentage reduction in cholesterol and the percentage reduction in CHD deaths during a 5-year trial (so that, for example, a 20% cholesterol reduction would lower CHD mortality in a 5-year trial by about 20%). The "expected" numbers of deaths and the statistical power of the current trials to detect such reductions in total mortality would then be as in Table 4.
Table 4: "Expected" numbers of deaths and power to detect effects on total mortality of current major HMG CoA reductase inhibitor trials 26 and of the Heart Protection Study(HPS)
5-6 yr av.
Estimated nos. of
|Primary prevention of MI|
|Overview of current trials|
|All primary prev.||16,000||Any||18%||350||255||285||320||12%|
*Power to detect a total mortality difference is calculated as the probability of achieving 2P<0.01, assuming no material effect on non-CHD deaths and an approximately one-to-one relationship between the percentage reduction in cholesterol and the percentage reduction in CHD deaths.
Secondary prevention of MI: Taken separately from each other, the current studies of secondary prevention among post-MI patients may well fail to demonstrate the sort of effects on total mortality that might realistically be expected (see Table 4). Combination of all their results in a systematic overview (or "meta-analysis") should allow such an effect on total mortality to be detected, but there is still a real chance that it might fail to do so. And, even if the overall results were significant, there would almost certainly be some major subgroup (e.g. males or females, young or old, moderately or extremely elevated cholesterol, hypertensive or not, etc) in which ambiguous results would lead to prolonged dispute, especially since an overview of several results is not as convincing to some clinicians as a single trial of adequate size.
Primary prevention of MI: The situation is much worse in the primary prevention of myocardial infarction among high-risk people, where the CHD death rates are lower and a larger proportion of deaths are from non-CHD causes. For, even an overview of the studies currently under way in primary prevention may fail to demonstrate the expected reductions in fatal CHD, let alone any effects on total mortality (see Table 4). Moreover, if any subgroups (e.g. male/female, middle-aged/old, etc) are to be analysed separately, to help determine which patients need treatment, then the power of the individual trials (and of any overview) is further reduced.
Non-CHD deaths: It might have been hoped that the current trials would at least be able to address reliably the suggestion that cholesterol-lowering therapy causes an increase in non-CHD deaths, but this too is uncertain. In total in all the current trials put together there are expected to be only about 600 non-CHD deaths (see Table 4), of which one-third might be from cancer and one-sixth from external causes such as accidents, violence or suicide. Even a meta-analysis of all these studies would have less than a one-in-three chance of detecting (or excluding) the sort of increases in deaths from all non-CHD causes (of about 20%), from cancer (of about 30%), and from external causes (of about 50%) suggested by some reviewers of the previous cholesterol-lowering trials 14-16. There is, therefore, a real risk of equivocal evidence about any effects of cholesterol lowering on non-CHD causes of death.
The present protocol is for a 6-year multicentre study large enough to demonstrate reliably the effects of cholesterol-lowering drug therapy on total mortality. At least 20,000 UK patients at high risk of CHD (e.g. those with a history of angina, coronary angioplasty [PTCA] or surgery [CABG], peripheral vascular disease, transient ischaemic attack [TIA] or non-haemorrhagic stroke, carotid endarterectomy, diabetes mellitus: irrespective of whether or not they have already suffered an MI) are to be randomised to compare simvastatin (40 mg daily) versus placebo and treated for at least 5 years. These groups all have a substantially increased absolute risk of coronary deaths and events compared with the general population, so even a moderate proportional reduction in risk might, in absolute terms, be worthwhile.
Total mortality and CHD mortality and morbidity in HPS: Among 20,000 high-risk patients there should be about 1575 CHD deaths (assuming 900 in the control group is reduced by about a quarter to 675 in the simvastatin group: see Table 4), plus similar numbers of non-fatal MIs, during about 5-6 years of follow-up. So, as well as providing unequivocal evidence about the overall effect of this particular cholesterol-lowering therapy on total mortality among these high-risk patients, a uniquely reliable assessment of its effects on CHD mortality and morbidity will also emerge. It is unlikely that a large reduction in cholesterol would reduce CHD mortality substantially in middle age and not at all in old age, or substantially in males and not at all in females, or just in patients with prior MI and not at all in others, etc but, of course, the exact size of the reduction in different subgroups may well be somewhat different. Separate analyses of the massively significant expected effects of treatment on the incidence of any CHD (fatal or not) in various subgroups should help identify any differences between the effects in different types of patient. Consequently, these subgroup findings for CHD incidence, along with the overall mortality findings, should provide reliable guidance as to what to expect in various specific circumstances. Moreover, by combining the results from this study with those from the other main randomised trials of cholesterol-lowering in a collaborative overview, it should be possible to assess directly the effects of lowering cholesterol on total mortality and on fatal CHD in some of the larger subgroups of interest. The Cholesterol-Lowering-Trialists' Collaboration has, therefore, been established in order to conduct a prospective systematic overview of the current and planned cholesterol-lowering trials after their results emerge.
Non-CHD mortality (and morbidity) in HPS: Among 20,000 high-risk patients there should also be about 1100 non-CHD deaths, with about 1100 new cancers, during about 5-6 years of follow-up (see Table 4). Hence, the Heart Protection Study will be able to assess reliably the effects of treatment not only on CHD mortality but also on particular non-CHD causes of death and on total cancer incidence. The results for non-CHD deaths will be particularly important because of the concerns expressed that such cholesterol-lowering therapy might have adverse effects on particular non-CHD causes (such as cancers, trauma, etc) 13-16. If there is no significant effect on non-CHD mortality as a whole (or, with appropriate allowance for multiple hypothesis testing, on particular non-CHD causes of death) with a substantial prolonged cholesterol reduction, then this should provide considerable reassurance that such cholesterol-lowering therapy is not hazardous. Consideration of the results from this trial along with the other main reductase inhibitor trials (in which a total of about 600 non-CHD deaths are expected) will allow even more detailed assessment of the effects of treatment on particular causes of death.
Health service issues: The proposed study would yield some of the information that is needed for cost-effectiveness analyses of cholesterol-lowering by these drugs in a wide range of patients at high risk of CHD. For, it would provide reliable estimates of the effects of such treatment on CHD mortality and morbidity, on non-CHD mortality, on non-fatal cancers, and on serious side-effects. This information, taken together with the absolute risk of CHD and non-CHD events, would yield an estimate of the absolute effectiveness and safety of such treatments in any particular risk group. Adequate economic and quality-of-life assessments using well validated methods are already being produced from several other large reductase inhibitor trials. This, in combination with the evidence from the Heart Protection Study should allow comprehensive cost-effectiveness analyses to be undertaken.
There is no longer any reasonable doubt that high plasma levels of LDL cholesterol are atherogenic and that lowering them can reduce the risk of CHD, but the specific cellular processes are less well understood. Recent studies suggest that LDL may be rendered toxic by oxidative modification that allows it to be taken up by macrophages in the artery walls 27,28. These macrophages, which are attracted to regions where oxidised LDL is being taken up, become loaded with cholesterol (and are then described as "foam cells" in the artery walls), leading to the development of "fatty streaks". Oxidised LDL can also be cytotoxic, which may play an additional role in the progression of the relatively benign coronary arterial fatty streaks to the ulcerated atherosclerotic plaques that are associated with clinical events.
In animal studies, antioxidants have been shown to slow the progression of atherosclerosis 29-31. Of the available antioxidants, dietary vitamins (for example, vitamins E and C and beta-carotene) have several advantages in particular, they occur naturally and appear to have few, if any, side-effects at doses that can increase plasma levels substantially 28. Vitamin E is the major antioxidant in LDL particles, and LDL does not become oxidatively modified in vitro until the associated vitamin E is first degraded 32,33. Oral supplementation of vitamin E substantially prolongs the resistance of LDL to oxidative damage 33. In a randomised placebo-controlled substudy of the pilot study for the Heart Protection Study, chronic daily dosing of 400 mg of vitamin E produced about a doubling of vitamin E levels in LDL cholesterol and of the lag-time to the onset of in vitro oxidation of LDL 34. Similar sized reductions in LDL oxidisability were seen in the presence and absence of simvastatin, and the vitamin supplement did not appear to interfere with the cholesterol-lowering effects of simvastatin.
Vitamin E may have other potentially protective effects, including inhibition of platelet activation 35, deactivation of protein kinase C 36, stabilisation of vascular tone 37 and, perhaps, infarct size limitation 38. Beta-carotene, which can also function as a fat-soluble antioxidant in certain special circumstances, is carried with vitamin E in the fatty cores of the LDL particles. Vitamin C is the major water-soluble antioxidant in the plasma, and it has been shown in vitro to regenerate oxidised vitamin E 39, and in animal studies to reverse atherosclerotic lesions 40.
In some large, well-conducted prospective epidemiological studies the intake of antioxidant vitamins was inversely associated with CHD incidence 41-43, in cross-population comparisons plasma vitamin levels were inversely associated with CHD 44, and in a case-control study low plasma levels were associated with an increased incidence of angina pectoris 45. The extent of atherosclerosis has also been reported to be related to levels of autoantibodies to oxidised LDL 46 and to the degree of LDL susceptibility to oxidative damage 47. Increased dietary consumption of antioxidant vitamins has also been found to be associated with lower rates of various cancers and of cataract 48.
Several editorials 49,50 and the participants in a NIH workshop 28 have, therefore, recently called for large-scale intervention trials of dietary antioxidants (in particular, vitamins E and C and beta-carotene) among high-risk individuals. But, even though there are many reasons to be hopeful, any protective effects may be only moderate. Consequently, if several trials are now undertaken, some may produce favourable results and some may fail to do so 51-53. Results are currently available from only one really large randomised trial 53. In this 2x2 factorial study with 6 years of treatment, a low daily dose of vitamin E was not shown to have protective effects and a standard daily dose of beta-carotene appeared, if anything, to have some possible adverse effects. These unpromising results are, however, so much at variance with the protective effects suggested from other types of evidence that the situation needs to be clarified by further large scale randomised trials. Indeed, it may be that only a systematic overview of all such studies will be able to provide reliable evidence about the effects of antioxidant vitamins 54.
What is needed is a larger amount of properly randomised evidence about really prolonged use of substantial doses of antioxidant vitamins. The Heart Protection Study among at least 20,000 individuals at high risk of CHD provides an excellent opportunity to help provide this by randomising high-risk individuals not only to compare simvastatin versus placebo but also to compare antioxidant vitamin supplementation versus placebo in a "2 x 2 factorial" design (see Section 2.3). The antioxidant vitamin regimen to be studied (600 mg of vitamin E, 250 mg of vitamin C and 20 mg of beta-carotene daily in 2 specially formulated capsules) is within the range that the NIH workshop considered likely to be safe and potentially effective 28.
The factorial design allows all patients to contribute fully to assessment of the separate effects of reductase inhibitor therapy and of antioxidant supplementation, without any material increase in sample size or non-drug cost beyond that required for a study that just assessed a reductase inhibitor. Such a study will provide important additional information about the combined effects of this cholesterol-lowering therapy and of antioxidant supplementation (which, it has been suggested, could well be additive or even synergistic 34,50). As well as increasing the scientific interest of the trial, inclusion of vitamin supplementation may well make the study of even greater interest to collaborating patients and their doctors.
HPS * Front Cover * Contents * Section 1 * Section 2 * Section 3 * Appendix * References * Back Cover