LDL-CHOLESTEROL AND CV RISK IN CKD: EPIDEMIOLOGY VS. CLINICAL TRIALS David Wheeler, London, UK Chairs: Maurice Laville, Lyon, France Erling B. Pedersen, Holstebro, Denmark Prof. David Wheeler Centre for Nephrology Royal Free Campus University College London Medical School London, UK Slide 1 So, these are my disclosures and I don't think any of them are particularly relevant to the talk. Slide 2
So, my topic is LDL cholesterol and cardiovascular risk in CKD and I was asked to compare the epidemiological data with the data that we've obtained recently from randomised controlled trials. Slide 3 So these are the latest data I could find from 2009 looking at the epidemiological association between here non-hdl cholesterol, which you can read as LDL cholesterol, and the hazards ratio for coronary heart disease. You'll see here adjusted data, adjusted for age and sex in circles and further adjusted for other cardiovascular risk factors in squares in 300.000 patients or people rather who didn't have vascular disease or didn't have overt vascular disease at the time these cholesterol measurements were taken. So, these individuals were then followed up and this is the hazards ratio for coronary heart disease events during the follow-up period. You can see as we're all well aware that the higher the baseline cholesterol concentration, that's the non-hdl or LDL cholesterol concentration, the greater the hazards ratio for a coronary heart disease event. Slide 4
If we look at these data more closely, we can calculate that for a 1 mmol reduction in LDL cholesterol, we would see or Slide 5 we would expect to see a 40% reduction in the hazards ratio for a coronary heart disease event, this is what the epidemiology tells us. Slide 6
The reality from clinical trials is not quite the same. So, what you're looking at here is a plot of data from all the major statin studies. The bigger the study, the larger the square and we're comparing here the proportional risk reduction for major coronary events in these trials against the reduction in LDL cholesterol that was achieved. You can see here a very nice relationship. The bigger the reduction in LDL cholesterol, the bigger the proportional risk of coronary heart disease events. Slide 7 Remember the 40% from the last slide that we expected from the epidemiology, if you look at the clinical trials from 1 mmol reduction in LDL cholesterol we actually see a 22% reduction in the risk of coronary heart disease events. So, something like half the reduction that we predicted from the epidemiological relationship which isn't perhaps surprising because of course, these trials treated patients for only around 5 years and the exposure to the LDL cholesterol in the epidemiological studies lasts a lifetime. But these trials, of course, encouraged us to use statins in patients who are at high risk of major coronary events and more recently, we've extended the use of stains into primary prevention based on the results of some of these studies. Slide 8
Now the situation in patients with kidney disease is slightly different and the epidemiology looks rather different. So, you're looking here at data from the UK Renal Registry that collects biochemical information from now all patients dialysed in the UK and you're looking at the relationship between serum cholesterol and the hazard of death. So this isn't specifically cardiovascular death but it is death. But what this shows us is a reverse relationship between serum cholesterol and the hazard of mortality. This holds both in haemodialysis patients and in peritoneal dialysis patients, the lower the LDL cholesterol, the higher the hazard of death. Slide 9 This is the wrong way round. O.K. you may say but we're looking at total mortality here but then let's look at cardiovascular death and here are data from 3 studies of dialysis populations again examining the relationship between total cholesterol and the relative risk for a cardiovascular death. Again, we see this reverse relationship. Certainly, in this study by Kilpatrick and this study by Degoulet the higher the cholesterol, the lower the risk of cardiovascular death and the lower the cholesterol, the higher the risk of cardiovascular death. This is the wrong way round. This might discourage us from lowering cholesterol in patients who are receiving dialysis. Slide 10
How about patients who are not on dialysis but have CKD? The data are not so good. This is a small study from Birmingham of patients who attended a nephrology clinic and had stages 3-5 of CKD but were not on dialysis. This is a relationship between the quintile of baseline cholesterol and the mortality both from cardiovascular disease in blue and non-cardiovascular disease in yellow. You'll see here no clear relationship between the higher levels of cholesterol and cardiovascular death. I accept that this is a small study but as I say, there are few studies that look at the relationship between cholesterol and cardiovascular disease events in non-dialysed CKD patients. Slide 11 Now we see this reverse relationship between cholesterol and mortality in other populations. As you can see from the scan here, the quality of the scan this is an old study from 1997 but looked at the relationship between total cholesterol and relative risk of coronary heart disease in a very elderly population. You'll see here a fairly flat relationship but with a higher risk among the elderly people who had the lower plasma cholesterol concentrations. It was speculated by these investigators that this was due to confounding. This was due to other factors, for example, malnutrition that both led to low cholesterol and a high risk of mortality. Slide 12
In fact, when these investigators corrected for these factors, they restored the expected relationship between cholesterol and the risk of cardiovascular or coronary heart disease deaths. So the reverse relationships certainly in this elderly population is probably due to confounding and when you correct for confounding, you restore the normal relationship. We can do this in dialysis populations. Slide 13 So you saw these data on a composite graph a few slides back. The relationship between cardiovascular mortality and blood cholesterol in a dialysis population. These investigators showed a fairly flat relationship with the suggestion of a reverse trend here, lower cholesterol, higher risk. This is their cohort overall. Slide 14
When they looked at those patients who had elevated inflammatory markers, high CRPs, their inflamed patients showed a similar relationship. But when they took those who had normal levels of inflammatory markers, so normal CRPs, Slide 15 they saw the expected relationship between total cholesterol and cardiovascular mortality in these dialysis patients. Suggesting that inflammation is confounding the relationship between cholesterol and cardiovascular events. Slide 16
So we argued some time ago that the only way to really untangle the relationship between cholesterol and vascular events in CKD patients in whom there are many confounding factors that make these epidemiological studies unreliable is to randomise patients and that means conducting randomised controlled trials, which of course we have done over the last few years. I'm going to show you the results of these trials and how they compare with the epidemiological predictions. Slide 17
So, if we look at statin studies from the 1990s, these are data from 3 studies that used pravastatin in different populations of patients with vascular disease and these data this is a post hoc analysis extracting from the datasets those patients who went into these trials with stage 3 CKD. So retrospective analysis of a large dataset from 3 statin studies. But, I think what's important here is that if you look at the patients with CKD and if you look at what happens during the trials, certainly in the secondary prevention setting here that is in those who'd already had a cardiovascular event there's a clear suggestion here that these patients benefitted from statin therapy. So lowering their cholesterol with pravastatin despite the presence of stage 3 CKD was associated with fewer further cardiovascular events and mortality during the follow-up period. Not quite such a clear relationship here in a primary prevention setting here and here. So, basically looking back through the datasets from the large statin studies, and this has been done in many of these studies, patients with stage 3 CKD seem to behave in the same way as the non-ckd population in these studies. Slide 18 Now in the late 1990s we set out to address the hypothesis that LDL lowering using a regime including a statin would reduce the risk of atherosclerotic vascular events in patients with CKD. I will come back to atherosclerotic vascular events in a minute. This was, of course, the origins of the SHARP study. Slide 19
The primary outcome in SHARP was vascular events driven by the process we know as atherosclerosis, the occlusion of arteries by lipid plaque in the intima. So, we chose our end points to reflect this pathology. Death due to occlusion of coronary arteries by atheroma. Non-fatal acute myocardial infarction due to plaque rupture. Non-haemorrhagic stroke and a revascularisation procedure, which is, of course, preformed to open up artery blocked by atherosclerosis. So, the primary outcome in SHARP was a composite of these atherosclerotic vascular events. We did of course, look at other vascular events in the subsidiary outcomes including death due to any cardiac cause and any form of stroke including haemorrhagic stroke. We had other outcomes including progression of kidney disease that I'm not going to talk about today. Slide 20 At the time that we wrote the SHARP protocol egfr reporting was not widely available and staging of CKD hadn't been invented. So, we took a fairly pragmatic approach and decided to recruit patients coming to clinic with elevated serum creatinine concentrations on two occasions. For men that meant a serum creatinine of 150 µmol/l and for women 130 µmol /l. We also recruited patients on dialysis and out of the 9.000 or so patients we recruited at the start of the study 6.000 had CKD and were not on dialysis and 3.000 had CKD and were receiving dialysis. Patients needed to be greater than 40 years of age to increase the risk and from the coronary artery perspective this was a primary prevention study. So patients didn't go into SHARP if they'd already had a myocardial infarction or contrary revascularisation for seizure, although we did allow other revascularisation procedures for example of peripheral arteries and this was not an exclusion criteria. But, from the cardiac perspective
this is a primary prevention study. Slide 21 You'll know the result of SHARP, the combination of simvastatin 20 mg and Ezetimibe 10 mg. This is a cholesterol absorption inhibitor that augments the action of the statin and reduces the risk of the atherosclerotic events endpoint that I've just shown you by 16.5%. So taking this result how does this compare with the statin studies in the general population? Slide 22 So, I'm taking you back to this graph where we plotted the statin studies and we plotted the
proportional reduction in risk against the reduction in LDL cholesterol. Slide 23 The average reduction of LDL cholesterol in SHARP was 0.84 mmol/l and we saw a 16.5% reduction in the major coronary events endpoint. As you'll see this is pretty much what we would have expected from the other statin studies. This result falls pretty much on this line drawn from this relationship based on the statin studies today. So in our patient population that was largely populated by patients who had CKD but who were not on dialysis the statin and Ezetimibe combination did what we expected it to do in the SHARP study. Slide 24
Of course, I can hear you telling me that there are two studies in dialysis patients, which didn't show a benefit of statins therapy. Here they are the 4D study from German investigators led by Christophe Wanner and the AURORA study led from Scandinavia by Bengt Fellström. Both of these studies used a statin, atorvastatin 20 mg in the 4D study, rosuvastatin 10 mg in the AURORA study. In both studies the LDL cholesterol was reduced by around about 0.9 mmol/l, so about the same as in SHARP but in neither study was there a clear benefit on cardiovascular events. So we have to work out why we've got different results from different studies. I would say up front that these studies both included a composite cardiovascular endpoint but these endpoints were slightly broader than in SHARP and included events that you could argue were not the direct result of the atherosclerotic process. Slide 25
When you look at the SHARP data, here's the result, a 16.5 % reduction in the risk of major atherosclerotic events favouring the lipid lowering intervention. You divide the patients between those who were not on dialysis and those who were on dialysis at the start of the study. You will see that the effect was less obvious in the patients who started the study on dialysis. Now, I should make a few points here. Firstly, if you do a statistical test to ask if these two are different, a test of heterogeneity, the answer is no. Statistically these two results are compatible. Secondly, I should point out that although these patients started the trial not on dialysis, about a third of these patients ended up in this dialysis category by the end of the study. So there's intermixing between the two populations here. I should also point out that SHARP was not powered to address the issue as to whether lipid-lowering intervention was beneficial in this smaller dialysis population. So the SHARP study was not really powered to separately analyse these two groups but there is this suggestion here that these patients who are on dialysis benefited less from the intervention. Slide 26 But let me just make another point. The dialysis patients were less likely to comply with their therapy. The compliance in this population was undoubtedly poorer and we collected data on compliance throughout the study. So, if you look at the mean LDL difference between the treated and the untreated group in the patients who started the trial not on dialysis it was nearly 1 mmol/l but in those patients who started the trial on dialysis the mean difference in LDL was much less, almost half that 0.6 mmol/l. This if you like lines up with the risk reduction that we saw in these two patient groups. Slide 27
So going back to this graph if you separately plot the patients who started the trial off dialysis, so these are the CKD patients not on dialysis at the start of the study, 22% reduction and if we look at those who started the trial on dialysis who had a lesser reduction in LDL cholesterol, we see a 10% reduction both in line with the data from the other statin studies. So you get a lesser reduction in LDL in the dialysis patients, you see a lesser benefit. So I would make the argument that in SHARP separate analysis of these two populations show that we get the expected result in both. Slide 28 One other point to make is that if we look at specific endpoints in the 4D, the AURORA and in the SHARP study, we see a homogeneity in terms of the outcome. So what I've plotted on this graph here and this was published in the SHARP paper is a composite plot of data looking at coronary revascularisation procedures in 4D, ALERT which was a transplant study, AURORA and SHARP. To the left of the neutrality line LDL lowering was better in terms of this outcome and to the right of this line LDL lowering was not beneficial. You can see clearly here that in 4D and in AURORA, although not statistically significant, the lipid lowering therapy that was
given in these studies produced a benefit in terms of coronary revascularisation. It's just that these results aren't statistically significant, whereas the SHARP result for this endpoint was statistically significant. If you do a meta-analysis and line these up, you see a clear benefit with a p value of 3 noughts and then with 4, so highly statistically significant. So, I guess the point I'm trying to make here is that if you take atherosclerotic endpoints from these trials that have involved dialysis patients and you plot these together, the data are pretty compatible. The test for heterogeneity here once again shows that these results are similar, they're not statistically different. So had 4D and AURORA looked at atherosclerotic endpoints, rather than broader cardiovascular endpoints, there's a possibility that the results of these studies might have been different. Slide 29 So, this is my summary slide. I hope I've shown you that the epidemiological relationship between total and LDL cholesterol and cardiovascular risk, particularly atherosclerotic events, is confounded in patients with advanced CKD. I've suggested that the only way to establish a causal role between LDL cholesterol and cardiovascular events in these patients because of this confounding is to conduct randomised controlled trials. I've shown you that when we conducted those trials, when we've lowered LDL and looked at the relevant endpoint that we would expect to change, then we see the expected reductions in relative risk associated with that endpoint. But I would concede that the value of LDL lowering in dialysis patients remains controversial despite the three trials to date that I've described and that could be a topic for discussion if we want. Thank you very much for listening to me. Slide 30
Chairman: Thank you Doctor Wheeler. The paper is now open for discussion. Please go to one of the mikes and start to identify yourself. Please Locatelli? Question: My question is from a practical point of view how many patients do we have to treat in order to save one in the different populations, in CKD not on dialysis and in patients on dialysis? Because this is for us as a clinician major information to be known. Prof. Wheeler: Ok so in SHARP it's 40 patients with CKD over 5 years to prevent one event. It's pretty similar to the data from the general population studies. So 30-40 patients, 5 year treatment prevents one major cardiovascular event in the context of SHARP. As I've shown you, that pretty much aligns with data from the other populations suggesting that that benefit is similar. In the dialysis population in SHARP you have to treat more patients to get that benefit. So, you're looking at 50-60 patients treated for 5 years to prevent one event in SHARP. So, in the dialysis population in SHARP more patients treated to prevent an event which you would expect from the relationship I showed you. Question: Is this number according to the guidelines something you suggest we do? Prof. Wheeler: Well is it worth doing? I guess is what you're asking me well that's the question. My own personal belief is that in younger patients this is well worth doing. So, if you have a young dialysis patient who is going to go on and get a transplant, whose life you know is going to be limited eventually by cardiovascular disease, then it makes complete sense to treat that patient's LDL cholesterol to reduce that risk. I think the situation is slightly different in the elderly dialysis patient who has a box full of pills and can't remember which to take in whom perhaps the atherosclerotic disease is a less important factor in the eventual outcome. So, you could argue that you should be targeting those patients, you know, who are most likely to benefit from this intervention. Question: Doctor Wheeler treatment is one thing monitoring is another thing. There has been some discussion about how to monitor treatment. If you decide to treat, would you use a 'fire and forget' strategy as has been suggested or treatment-to-target? Please comment on that. Prof. Wheeler: So 'fire and forget'. So you'll be aware that we've had a discussion in the production of the latest KDIGO guidelines about the value of repeated measurements of LDL cholesterol after you've treated. There is a view that whatever your risk of coronary heart disease, you will have a lower risk if you lower LDL cholesterol. Therefore, a number of people have suggested that if you simply put everybody just on the effective therapy on the statin, know you've done what you can, then you can forget about that risk factor and move on to worrying about the others. So, there is a suggestion that one of our approaches should be simply to say this is a high risk patient, I'm going to put this patient on a statin in a reasonable dose that is tolerated by the patient and is safe and then I'm not going to worry about that risk factor anymore because I've done what I can about it. There's another review from perhaps a slightly more obsessional people who argue that what you want to do is know that you've reduced your LDL and know that if it's only so low, you might be able to reduce it further by increasing the dose of the statin of adding in Ezetimibe. So the big debate now is after you've started the treatment do you need to monitor the effect and titrate up or do you simply say I've treated it, I've done my best, I've given a reasonable dose of statin, I'm a busy nephrologists I'm going to move on and do something else a little bit more
interesting. I think this is the debate that we've had and it's a debate that will be reflected in the forthcoming KDIGO guideline on the management of dyslipidaemia in CKD which we hope will be published in June or July as a Kidney International supplement. Chairman: Please. Question: --- Manheim Germany. We have shown in the 4D study, I guess that you are aware of that publication, that the LDL cholesterol level at baseline determines the clinical outcome in other words that you get a highly significant reduction in events if you look into the fourth quartile which starts at an LDL cholesterol of 140, so have you done the same things in SHARP? Prof. Wheeler: Yes. Question: And in particular what happens to the dialysis patients? Do they show the same phenomenon, a larger relative risk reduction at high baseline LDL cholesterol levels? Prof. Wheeler: Yes so we have to done that analysis in SHARP, it's actually in the paper but it's hidden in then paper. We see the same trend, so the higher the start in LDL cholesterol, the greater if you like there is a trend towards a greater benefit of statin therapy. We have not done that separately for the dialysis population and the CKD populations in SHARP. I think our argument would be that the numbers are smaller and that the trial wasn't powered to do that but we certainly see the expected trend across the starting LDL cholesterol categories. As in the general population your risk is higher if your starting LDL is higher and the proportional benefit is greater in those patients who have the higher starting LDL cholesterol. Question: One could easily increase the power to look at dialysis patients separately if one lumped together AURORA, SHARP and dialysis subgroup and the 4D study. Prof. Wheeler: So there have been plans for a patient level meta-analysis of the dialysis data in the relevant studies. I'm not quite sure where that's got to but it involves quite a lot of work as you'll appreciate but that's a good idea and it has been thought through and suggested and it may happen. Chairman: A last question for Doctor Wheeler. Question: Chris from Bristol. David should we risk stratify the CKD patients or should everybody other than those with awful prognosis be on statins? Prof. Wheeler: So should we risk stratify the CKD patients? I think it depends on your philosophy, so I think the message from the general population is if that whatever your starting LDL cholesterol, you will benefit by reducing that to a lower level with a statin. So if we were going to ration our care, if we had a budget which we do in the NHS and we wanted to target the therapy perhaps at the most relevant patients, then I think yes, it's very reasonable to risk stratify. How we risk stratify is slightly more complicated because of course the epidemiological data as I've shown you don't always tell us the patients at highest risk. But my view is that here we have a therapy that's effective in patients with CKD that reduces an endpoint that's relevant. The therapy is safe and it's cheap. I would argue that this is something we should be spending some time on rather than giving something like folic acid, which doesn't have an evidence base to support it. So I'm not in favour of risk stratification, I'm in favour of getting the effect of therapy to the maximum number of patients who will see the greatest benefit. Sorry the maximum number of patients, all of whom will benefit I believe. Chairman: Ok thank you very much Doctor Wheeler.